diff --git a/.gitattributes b/.gitattributes
index bd22570f3bfab06965e86ab3fc867e5ba5b4959b..0c2c53eca7a8a0dc9ad80b73e8fe6fb228bdce96 100644
--- a/.gitattributes
+++ b/.gitattributes
@@ -35,3 +35,1346 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.mp3 filter=lfs diff=lfs merge=lfs -text
 *.ogg filter=lfs diff=lfs merge=lfs -text
 *.wav filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.wuu_Hani-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cat-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ell-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ota.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-nld.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dan-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-bul.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-nor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-toki.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-ckb.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cmn_Hans-yue_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hun-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ben-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.swe-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-lzh_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-lzh_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lad-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ces-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.aze-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ber_Latn-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-ina.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hrv-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ber_Latn-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-tat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-vie.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ina-lad.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.por-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lad_Latn-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lit-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-nor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-hun.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-jpn_Kana.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nld-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cor-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-ile_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lit-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-kur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-msa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-ina.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nor-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-nor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hbs-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.run-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tlh-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-nld.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-fin.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-wuu_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ota_Arab.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.mkd-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cor-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-fin.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-fkv.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-uig.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jbo_Latn-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tur-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-toki.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-ron.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-toki.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ell-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-lfn_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lat_Latn-lfn_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Kana-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-ron.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nld-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-ina.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ron-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kaz-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tur-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ina.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-pms.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-srp_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-kor_Hang.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-yue_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-nob.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn-nds.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-slv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lad-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.pol-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cmn_Hant-yue_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-heb.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Kana-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hbs-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nor-nor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.pol-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ina-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-nno.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.khm-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-lfn_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Kana-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lat-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.wuu-yue_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-ber.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-gle.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nob-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tur-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.glg-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dtp-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-fin.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-hbs.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nds-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nob-swe.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.pol-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-gle.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-tuk_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-vie.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.grn-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn-vie.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-kzj.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-tgl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-tgl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.zsm_Latn-ind.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-kor_Hang.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dsb-hsb.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina_Latn-lad_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nds-nld.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-hun.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.srp_Latn-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tur-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.isl-ita.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ber-fra.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-bos_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jbo_Latn-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nno-nob.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dan-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lfn_Latn-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-toki.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-lad_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-lit.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-lzh_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-kur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-fin.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ber-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.arq-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.bel-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-cmn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-fao.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-tuk.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ron-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-lzh_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-fas.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-est.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cat-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hani-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-toki.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-fin.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ell-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lat-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-orv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-jbo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ido-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nor-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.wuu-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ber_Latn-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-kmr_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hani-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.wuu-yue_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cmn_Hans-wuu_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-yue_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-kaz_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-fin.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-fra.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nob-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tur-uzb.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ber_Latn-fra.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lfn_Cyrl-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nds-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nob-nno.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-jbo_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-lad.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eus-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.glg-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dan-nob.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-tgl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kab-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lad-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.orv-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nor-nor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jbo-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-kor_Hang.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-ita.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hun-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-kab.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dtp-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-pcd.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-lit.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-isl.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.por-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-jbo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-tam.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ukr-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eus-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-hun.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.pol-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-uig.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-ita.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-hin.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-ron.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.afr-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lit-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-tuk.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-zsm_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-swg.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-msa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-que.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lad-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.msa-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-hin.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-lad.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-nds.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-pes.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tat-vie.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.vie-vie.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-lav.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ber-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.afr-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-srp_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ido-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.srp_Cyrl-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn-msa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hani-vie.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-heb.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-nor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-kor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-zza.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-slv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lfn_Cyrl-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.msa-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-swg.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hun-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-dsb.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-fas.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ido_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ell-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-isl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kaz_Cyrl-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.bul-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Kana-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lad-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-vol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-toki.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ind.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-lat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-heb.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-ind.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hani-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lav-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.por-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cor-deu.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cmn_Hant-yue_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-epo.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.grn-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-slv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hsb-slv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-ron.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.afr-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.avk-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dtp-zsm_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-urd.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-hun.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-msa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.slv-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.yid-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.avk-fra.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.bos_Latn-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-ido.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ell-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hani-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-hun.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hani-ind.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-lzh_Kana.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bre-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.chv-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-uig_Arab.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ara-heb.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eus-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dan-nld.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-heb.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-msa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-lfn_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-lzh_Bopo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-hun.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-nno.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ota-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ell-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-hun.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-ind.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-ind.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-nds.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-jpn_Kana.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nld-ron.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.avk-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-ile.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kor_Hang-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.swe-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cat-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ron.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-vol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-swe.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ota-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-tat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.bre-fra.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.bel-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.bel-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-hbs.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-lad.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.pol-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cmn_Hans-yue_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lfn_Cyrl-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.chv-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ell-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hbs-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tlh_Latn-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-vol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-toki.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-isl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-jbo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hbs-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kab-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jbo-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ron-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ell-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ber-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-mlt.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-yue_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-ron.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kab-kab.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cmn_Hant-wuu.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-kab.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.egl-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-run.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-nor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-heb.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ina_Latn-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-grc.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-hbs.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eus-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hye-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hun-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-nno.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.por-toki.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-run.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina_Latn-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jbo_Latn-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lfn-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lfn_Cyrl-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-glg.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-vie.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ukr-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lat_Latn-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ber-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-ina.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.est-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-ina.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hani-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-nds.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nds-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-lzh_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-ron.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-oci.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jbo-swe.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-sah.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lad_Latn-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hani-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lfn_Latn-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nob-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cbk-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-nld.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dtp-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-kor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-frr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-kor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-kor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cor-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-heb.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lat-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.afr-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ell-ell.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-war.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ber_Latn-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-msa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-nds.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.srp_Cyrl-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-slv.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ina-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hani-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-lit.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-swe.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-gos.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-tlh_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lfn-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.uig-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-fas.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-mkd.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ina_Latn-lad_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-nor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tlh-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ell-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-zlm_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-nov.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kor_Hang-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kzj_Latn-zsm_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-tgl.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-nno.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-ber_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-est.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-kur_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-tgl.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.kaz-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-ido.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-ile.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-srp_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tlh_Latn-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dan-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hani-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.pol-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ces-deu.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.gcf-gcf.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.swe-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dan-epo.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-toki.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ell-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-kor_Hang.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat_Latn-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nor-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-slv.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-wuu.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-zsm_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.tlh-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-tlh_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Kana-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-nds.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-tat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-lat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-yue_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.yue_Hans-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-hbs.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-eus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hbs-nor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-heb.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.pol-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ron-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-max_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ara-epo.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-ina.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hrv-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ben-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-lfn_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-srp_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-ita.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-kor_Hang.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-ron.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-jpn_Kana.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.aze-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-mar.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-lzh_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-tat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ara-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cmn_Hant-lzh_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-nno.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-gos.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-ido.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.wuu-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-yue_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-vie.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-lat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Kana-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hbs-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ary-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-heb.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cat-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-yue_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.aze_Latn-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cor-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-tha.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-nob.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-nor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ina-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ceb-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-gos.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.grn-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-vie.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-fkv.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-uig_Arab.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-lfn_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-yue_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-kor_Hang.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jbo-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nor-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-ita.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-vie.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nld-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.aze-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cat-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-kur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-srp_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina_Latn-tlh_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-ile.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-nds.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-hun.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-lzh_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-ell.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bua_Cyrl-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-ltz.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-hrv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.gla-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.bul-ita.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-fas.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-tzl.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-lad_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-hye.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-kha.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-hbs.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dan-fin.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-jbo.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-toki.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ber-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-nor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-pam.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.msa-msa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-lzh_Hang.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-tat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-jpn_Kana.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hun-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ina-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lad-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-uig_Arab.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-epo.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.yue_Hant-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-vie.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-khm.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.pol-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ara-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hrv-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-nno.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-toki_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-run.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-nor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-slv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-swe.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.swe-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.afb-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-lad.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-dan.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-est.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-lit.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ltz-nld.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-yue_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ido-ina.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ina-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cor-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-oci.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-jpn_Kana.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-yue_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-hun.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-lit.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-kor_Hang.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-ind.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nld-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-got.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-fin.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-jpn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dsb-slv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nor-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-nds.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lad_Latn-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-yue_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.srp_Latn-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ces-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-fra.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hun-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-prg.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kor_Hang-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lat-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hbs-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ara-deu.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cat-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.afr-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fry-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lfn_Latn-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ber-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ell-swe.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hin-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cat-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-uzb_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-mon.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hani-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nld-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.isl-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-ind.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ces-epo.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-lad.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.afr-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-slv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hrv-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ber_Latn-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ces-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-ido.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nds-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.slv-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-fry.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat-nor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-srp_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina-lad.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.isl-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cmn_Hans-wuu.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dan-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nor-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ces-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hun-jpn_Kana.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-ido.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-fin.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-kor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-jpn_Kana.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lad-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-uzb.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-vie.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.awa-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.srp_Latn-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ukr-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ces-hun.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ell-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-hbs.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-toki.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bre-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ota_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Kana-ind.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cat-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dan-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-lit.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cha-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-mal.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-heb.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lfn_Latn-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-nst.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.chm-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-lat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-bos_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-hoc.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-nds.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-lad.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nld-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-jbo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-jpn_Kana.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-tat_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-eus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ido_Latn-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-lad_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-ilo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-toki_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-pms.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.uig_Arab-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-srp_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tlh_Latn-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-fra.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lad_Latn-lat_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dan-nor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kzj-msa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lfn-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-ell.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-lfn_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-isl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-kaz.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-ron.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tat-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lfn-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-ind.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dtp-msa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ido-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-hrv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-swe.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-vie.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hin-urd.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kat-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lfn-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tur-uig.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-hye.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hun-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.yue_Hans-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cym-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-toki.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lat-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-ell.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-lit.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ber_Latn-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cmn_Hans-wuu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-eus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-ita.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lfn-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-bos_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-jpn_Kana.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-tgl.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ina-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lfn-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kor-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-nds.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-ina.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.por-ron.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ara-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-wuu.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.srp_Cyrl-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.crh-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-zsm_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-hrv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-tgl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ell-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jbo-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hrv-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-toki.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jbo-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nld-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-hun.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-run.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-toki.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.zho-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.arq-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ceb-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-sqi.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-kur_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-toki.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-kur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ido_Latn-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cbk-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.srp_Latn-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-mar.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ido-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-srp_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-kab.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jbo_Latn-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cor-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-ile.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-lat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-toki.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hani-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nld-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-swe.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-toki.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-srp_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Kana-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.chm-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-vie.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-glg.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ind-zsm_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.slv-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ara-ber.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.aze-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-ron.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-lad_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ido-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ara-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina-lat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-vol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-swa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-yue_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-hrv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hira-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.wuu-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bul-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-nor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-hun.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-tgl_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-lad.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-xal.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-fas.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.bul-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lit-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lad_Latn-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-hrv.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.pol-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-jpn_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hani-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-xal.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.zho-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-yue_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.chv-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-vol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-fin.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-jbo_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fas-fra.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nld-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-lad_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-hoc_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-ido.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-nds.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ell-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-lit.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-pam.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-uig.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cmn_Hant-wuu_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hani-yue_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-hun.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.grn-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ber-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-ita.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lad-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-ile.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nor-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lfn_Latn-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-slv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ido-ina.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.nld-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-tzl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-ina.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-msa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-jav.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-kab.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-ita.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ara-ber_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ido-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-lat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lad_Latn-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-nob.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.heb-nld.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ido-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hani-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-uig_Arab.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.gos-nld.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-swe.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-tel.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hun-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.isl-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lit-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cat-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-pol.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-wuu_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.pol-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ina_Latn-lfn_Cyrl.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-por.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-tat.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ell-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hrv-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lat-yid.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ido.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.uig_Arab-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fin-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-swe.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-heb.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-mri.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-kab.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.gos-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lfn-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cor-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.dan-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.por-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.srp_Latn-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-ile.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn_Hira-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kor-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-jpn_Hira.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lzh_Hani-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.pol-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-nor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ita-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-ido.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-mar.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-lad.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-tlh.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.kor-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-kha.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.vie-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.tur-uig.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.pol-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bel-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-hrx.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-hrx.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-hsb.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-gsw.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.glg-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hbs-nor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.arz-eng.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ell-nld.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-ina.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.lzh_Hani-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ltz.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-mkd.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.vie-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-yue_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Hani-nld.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.kab-kab.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-lit.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.por-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-nor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-fra.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-nld.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.jpn_Kana-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-nld.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.hbs-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-uig.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-wuu_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hun-rus.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-deu.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.srp_Cyrl-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ces-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-kor.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.afr-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-eng.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-kat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-yid.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-yue_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.srp_Latn-nob.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.yue_Hant-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dan-swe.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.heb-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ita-pol.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat_Latn-lfn_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-gla.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-mon.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-rom.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-gcf.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.run-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.slv-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.swe-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.cmn_Hans-lzh_Hani.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ina-lat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.wuu_Hani-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.bua-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-mya.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.jpn-mar.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.rus-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-fas.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-spa.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hun-ita.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ain-fin.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-heb.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ukr-cmn_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.spa-yue_Hans.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.tur-tur.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ilo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.spa-vie.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.rus-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cat-epo.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-wuu.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.ara-fra.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-nor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.hbs-rus.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.cat-deu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fin-zho.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ina-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.tur-ukr.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.ara-heb.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.epo-hrv.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.tur-cmn_Hant.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-nob.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-ile.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-oci.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.fra-swe.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-wuu.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.dtp-jpn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.eng-tzl_Latn.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.nds-por.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.deu-tat.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.fra-spa.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.deu-kor.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-lfn.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-tur.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.eng-zho.tsv filter=lfs diff=lfs merge=lfs -text
+dev/tatoeba-dev.epo-nds.tsv filter=lfs diff=lfs merge=lfs -text
+test/tatoeba-test.lat-lat.tsv filter=lfs diff=lfs merge=lfs -text
diff --git a/README.md b/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..9de3764e03e890097a9251daa90376f32b4941be
--- /dev/null
+++ b/README.md
@@ -0,0 +1,284 @@
+---
+annotations_creators:
+- no-annotation
+language_creators:
+- crowdsourced
+languages:
+- af
+- ar
+- az
+- be
+- bg
+- bn
+- br
+- bs
+- ca
+- ch
+- cs
+- cv
+- cy
+- da
+- de
+- el
+- en
+- eo
+- es
+- et
+- eu
+- fa
+- fi
+- fo
+- fr
+- fy
+- ga
+- gd
+- gl
+- gn
+- he
+- hi
+- hr
+- hu
+- hy
+- ia
+- id
+- ie
+- io
+- is
+- it
+- ja
+- jv
+- ka
+- kk
+- km
+- ko
+- ku
+- kw
+- la
+- lb
+- lt
+- lv
+- mi
+- mk
+- ml
+- mn
+- mr
+- ms
+- mt
+- my
+- nb
+- nl
+- nn
+- no
+- oc
+- pl
+- pt
+- qu
+- rn
+- ro
+- ru
+- sh
+- sl
+- sq
+- sr
+- sv
+- sw
+- ta
+- te
+- th
+- tk
+- tl
+- tr
+- tt
+- ug
+- uk
+- ur
+- uz
+- vi
+- vo
+- yi
+- zh
+
+licenses:
+- cc-by-2.0
+multilinguality:
+- translation
+pretty_name: The Tatoeba Translation Challenge
+size_categories:
+- unknown
+source_datasets:
+- original
+task_categories:
+- conditional-text-generation
+task_ids:
+- machine-translation
+---
+
+# Dataset Card for [Dataset Name]
+
+## Table of Contents
+- [Table of Contents](#table-of-contents)
+- [Dataset Description](#dataset-description)
+  - [Dataset Summary](#dataset-summary)
+  - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
+  - [Languages](#languages)
+- [Dataset Structure](#dataset-structure)
+  - [Data Instances](#data-instances)
+  - [Data Fields](#data-fields)
+  - [Data Splits](#data-splits)
+- [Dataset Creation](#dataset-creation)
+  - [Curation Rationale](#curation-rationale)
+  - [Source Data](#source-data)
+  - [Annotations](#annotations)
+  - [Personal and Sensitive Information](#personal-and-sensitive-information)
+- [Considerations for Using the Data](#considerations-for-using-the-data)
+  - [Social Impact of Dataset](#social-impact-of-dataset)
+  - [Discussion of Biases](#discussion-of-biases)
+  - [Other Known Limitations](#other-known-limitations)
+- [Additional Information](#additional-information)
+  - [Dataset Curators](#dataset-curators)
+  - [Licensing Information](#licensing-information)
+  - [Citation Information](#citation-information)
+  - [Contributions](#contributions)
+
+## Dataset Description
+
+- **Homepage:** https://github.com/Helsinki-NLP/Tatoeba-Challenge/
+- **Repository:** https://github.com/Helsinki-NLP/Tatoeba-Challenge/
+- **Paper:** [The Tatoeba Translation Challenge – Realistic Data Sets for Low Resource and Multilingual MT](https://aclanthology.org/2020.wmt-1.139/)
+- **Leaderboard:**
+- **Point of Contact:** [Jörg Tiedemann](mailto:jorg.tiedemann@helsinki.fi)
+
+### Dataset Summary
+
+The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.
+
+### Supported Tasks and Leaderboards
+
+The translation task is described in detail in the [Tatoeba-Challenge repository](https://github.com/Helsinki-NLP/Tatoeba-Challenge) and covers various sub-tasks with different data coverage and resources. [Training data](https://github.com/Helsinki-NLP/Tatoeba-Challenge/blob/master/data/README.md) is also available from the same repository and [results](https://github.com/Helsinki-NLP/Tatoeba-Challenge/blob/master/results/tatoeba-results-all.md) are published and collected as well. [Models](https://github.com/Helsinki-NLP/Tatoeba-Challenge/blob/master/results/tatoeba-models-all.md) are also released for public use and are also partially available from the [huggingface model hub](https://huggingface.co/Helsinki-NLP).
+
+
+### Languages
+
+The data set covers hundreds of languages and language pairs and the original data sets are organized by ISO-639-3 languages. In this copy of the test and development data, the subsets are provided with BCP-47 codes. The current release covers the following languages: Afrikaans, Arabic, Azerbaijani, Belarusian, Bulgarian, Bengali, Breton, Bosnian, Catalan, Chamorro, Czech, Chuvash, Welsh, Danish, German, Modern Greek, English, Esperanto, Spanish, Estonian, Basque, Persian, Finnish, Faroese, French, Western Frisian, Irish, Scottish Gaelic, Galician, Guarani, Hebrew, Hindi, Croatian, Hungarian, Armenian, Interlingua, Indonesian, Interlingue, Ido, Icelandic, Italian, Japanese, Javanese, Georgian, Kazakh, Khmer, Korean, Kurdish, Cornish, Latin, Luxembourgish, Lithuanian, Latvian, Maori, Macedonian, Malayalam, Mongolian, Marathi, Malay, Maltese, Burmese, Norwegian Bokmål, Dutch, Norwegian Nynorsk, Norwegian, Occitan, Polish, Portuguese, Quechua, Rundi, Romanian, Russian, Serbo-Croatian, Slovenian, Albanian, Serbian, Swedish, Swahili, Tamil, Telugu, Thai, Turkmen, Tagalog, Turkish, Tatar, Uighur, Ukrainian, Urdu, Uzbek, Vietnamese, Volapük, Yiddish, Chinese
+
+
+## Dataset Structure
+
+### Data Instances
+
+Data instances are given as translation units in TAB-separated files with six columns: source and target language BCP-47 codes, source and target language ISO-639-3 codes, source language text and target language text. Note that we do not imply a translation direction and consider the data set to be symmetric and to be used as a test set in both directions. Language-pair-specific subsets are only provided under the label of one direction using sorted ISO-639-3 language IDs.
+
+Some subsets contain several sub-languages or language variants. They may refer to macro-languages such as Serbo-Croatian languages that are covered by the ISO code `hbs`. Language variants may also include different writing systems and in that case the ISO15924 script codes are attached to the language codes. Here are a few examples from the English to Serbo-Croation test set including examples for Bosnian, Croatian and Serbian in Cyrillic and in Latin scripts:
+
+```
+en	bs_Latn	eng	bos_Latn	Children are the flowers of our lives.	Djeca su cvijeće našeg života.
+en	hr	eng	hrv	A bird was flying high up in the sky.	Ptica je visoko letjela nebom.
+en	sr_Cyrl	eng	srp_Cyrl	A bird in the hand is worth two in the bush.	Боље врабац у руци, него голуб на грани.
+en	sr_Latn	eng	srp_Latn	Canada is the motherland of ice hockey.	Kanada je zemlja-majka hokeja na ledu.
+```
+
+There are also data sets with sentence pairs in the same language. In most cases, those are variants with minor spelling differences but they also include rephrased sentences. Here are a few examples from the English test set:
+
+```
+en      en      eng     eng     All of us got into the car.     We all got in the car.
+en      en      eng     eng     All of us hope that doesn't happen.     All of us hope that that doesn't happen.
+en      en      eng     eng     All the seats are booked.       The seats are all sold out.
+```
+
+
+### Data Splits
+
+Test and development data sets are disjoint with respect to sentence pairs but may include overlaps in individual source or target language sentences. Development data should not be used in training directly. The goal of the data splits is to create test sets of reasonable size with a large language coverage. Test sets include at most 10,000 instances. Development data do not exist for all language pairs.
+
+To be comparable with other results, models should use the training data distributed from the [Tatoeba MT Challenge Repository](https://github.com/Helsinki-NLP/Tatoeba-Challenge/) including monolingual data sets also listed there.
+
+
+## Dataset Creation
+
+### Curation Rationale
+
+
+The Tatoeba MT data set will be updated continuously and the data preparation procedures are also public and released on [github](https://github.com/Helsinki-NLP/Tatoeba-Challenge/). High language coverage is the main goal of the project and data sets are prepared to be consistent and systematic with standardized language labels and distribution formats.
+
+
+### Source Data
+
+#### Initial Data Collection and Normalization
+
+The Tatoeba data sets are collected from user-contributed translations submitted to [Tatoeba.org](https://tatoeba.org/) and compiled into a multi-parallel corpus in [OPUS](https://opus.nlpl.eu/Tatoeba.php). The test and development sets are incrementally updated with new releases of the Tatoeba data collection at OPUS. New releases extend the existing data sets. Test sets should not overlap with any of the released development data sets.
+
+
+#### Who are the source language producers?
+
+The data sets come from [Tatoeba.org](https://tatoeba.org/), which provides a large database of sentences and their translations into a wide varity of languages. Its content is constantly growing as a result of voluntary contributions of thousands of users.
+The original project was founded by Trang Ho in 2006, hosted on Sourceforge under the codename of multilangdict. 
+
+
+### Annotations
+
+#### Annotation process
+
+Sentences are translated by volunteers and the Tatoeba database also provides additional metadata about each record including user ratings etc. However, the metadata is currently not used in any way for the compilation of the MT benchmark. Language skills of contributors naturally vary quite a bit and not all translations are done by native speakers of the target language. More information about the contributions can be found at [Tatoeba.org](https://tatoeba.org/).
+
+
+#### Who are the annotators?
+
+### Personal and Sensitive Information
+
+For information about handling personal and sensitive information we refer to the [original provider](https://tatoeba.org/) of the data. This data set has not been processed in any way to detect or remove potentially sensitve or personal information.
+
+## Considerations for Using the Data
+
+### Social Impact of Dataset
+
+The language coverage is high and with that it represents a highly valuable resource for machine translation development especially for lesser resourced languages and language pairs. The constantly growing database also represents a dynamic resource and its value wil grow further.
+
+
+### Discussion of Biases
+
+The original source lives from its contributors and there interest and background will to certain subjective and cultural biases. Language coverage and translation quality is also biased by the skills of the contributors.
+
+
+### Other Known Limitations
+
+The sentences are typically quite short and, therefore, rather easy to translate. For high-resource languages, this leads to results that will be less useful than more challenging benchmarks. For lesser resource language pairs, the limited complexity of the examples is actually a good thing to measure progress even in very challenging setups.
+
+
+## Additional Information
+
+### Dataset Curators
+
+The data set is curated by the University of Helsinki and its [language technology research group](https://blogs.helsinki.fi/language-technology/). Data and tools used for creating and using the resource are [open source](https://github.com/Helsinki-NLP/Tatoeba-Challenge/) and will be maintained as part of the [OPUS ecosystem](https://opus.nlpl.eu/) for parallel data and machine translation research.
+
+
+### Licensing Information
+
+The data sets are distributed under the same licence agreement as the original Tatoeba database using a
+[CC-BY 2.0 license](https://creativecommons.org/licenses/by/2.0/fr/). More information about the terms of use of the original data sets is listed [here](https://tatoeba.org/eng/terms_of_use).
+
+
+### Citation Information
+
+If you use the data sets then, please, cite the following paper: [The Tatoeba Translation Challenge – Realistic Data Sets for Low Resource and Multilingual MT](https://aclanthology.org/2020.wmt-1.139/)
+
+```
+@inproceedings{tiedemann-2020-tatoeba,
+    title = "The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}",
+    author = {Tiedemann, J{\"o}rg},
+    booktitle = "Proceedings of the Fifth Conference on Machine Translation",
+    month = nov,
+    year = "2020",
+    address = "Online",
+    publisher = "Association for Computational Linguistics",
+    url = "https://aclanthology.org/2020.wmt-1.139",
+    pages = "1174--1182",
+}
+```
+
+
+[More Information Needed]
+
+### Contributions
+
+Thanks to [@jorgtied](https://github.com/jorgtied) and [@Helsinki-NLP](https://github.com/Helsinki-NLP) for adding this dataset.
+Thanks also to [CSC Finland](https://www.csc.fi/en/solutions-for-research) for providing computational resources and storage space for the work on OPUS and other MT projects.
diff --git a/dataset_info.json b/dataset_info.json
new file mode 100644
index 0000000000000000000000000000000000000000..3776a7ef2088c96728dea64ec4b8ea50db4dc22c
--- /dev/null
+++ b/dataset_info.json
@@ -0,0 +1,43782 @@
+{
+   "epo-glg" : {
+      "splits" : {
+         "test" : {
+            "num_bytes" : 24476,
+            "name" : "test",
+            "num_examples" : 329,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 24476,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "epo-glg",
+      "dataset_size" : 329
+   },
+   "gla-spa" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 19141,
+            "name" : "test",
+            "num_examples" : 289,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 19141,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "gla-spa",
+      "dataset_size" : 289
+   },
+   "kzj-msa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "kzj-msa",
+      "dataset_size" : 370,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 45722,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 45722,
+            "num_examples" : 370,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "kab-spa" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 883,
+      "config_name" : "kab-spa",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 70630,
+            "num_examples" : 883,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 70630,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "jpn_Hira-rus" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 18016,
+      "config_name" : "jpn_Hira-rus",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 9008,
+            "num_bytes" : 1119177
+         },
+         "test" : {
+            "num_bytes" : 1119177,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 9008,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 2238354,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "rus-spa" : {
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2100744,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 10506,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1050372
+         },
+         "validation" : {
+            "num_bytes" : 1050372,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10506
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 21012,
+      "config_name" : "rus-spa",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "eng-est" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 175534,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 1359,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 87767
+         },
+         "test" : {
+            "num_bytes" : 87767,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1359,
+            "name" : "test"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 2718,
+      "config_name" : "eng-est",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-gle" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 3826,
+      "config_name" : "eng-gle",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 237014,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1913,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 118507
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1913,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 118507
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "ain-fin" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "ain-fin",
+      "dataset_size" : 206,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 14201,
+            "name" : "test",
+            "num_examples" : 206,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 14201,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "lat-por" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1252072,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 5001,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 626036
+         },
+         "test" : {
+            "num_bytes" : 626036,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 5001,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 10002,
+      "config_name" : "lat-por",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fin-rus" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "fin-rus",
+      "dataset_size" : 7286,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 323616,
+            "num_examples" : 3643,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 3643,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 323616
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 647232,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "deu-swg" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "deu-swg",
+      "dataset_size" : 3046,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1523,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 166882
+         },
+         "validation" : {
+            "num_bytes" : 166882,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1523
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 333764,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null
+   },
+   "bel-pol" : {
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 26236,
+            "name" : "test",
+            "num_examples" : 287,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 26236,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 287,
+      "config_name" : "bel-pol",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ina-rus" : {
+      "config_name" : "ina-rus",
+      "dataset_size" : 2906,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 351908,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1453,
+            "num_bytes" : 175954
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1453,
+            "num_bytes" : 175954
+         }
+      },
+      "task_templates" : null
+   },
+   "lfn_Latn-yid" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 163694,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 81847,
+            "num_examples" : 898,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 81847,
+            "name" : "test",
+            "num_examples" : 898,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "dataset_size" : 1796,
+      "config_name" : "lfn_Latn-yid",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "pol-ukr" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "pol-ukr",
+      "dataset_size" : 5038,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2519,
+            "num_bytes" : 199615
+         },
+         "validation" : {
+            "num_bytes" : 199615,
+            "name" : "validation",
+            "num_examples" : 2519,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 399230
+   },
+   "epo-nor" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "epo-nor",
+      "dataset_size" : 6316,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 3158,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 313130
+         },
+         "validation" : {
+            "num_bytes" : 313130,
+            "name" : "validation",
+            "num_examples" : 3158,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 626260
+   },
+   "est-rus" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 613,
+      "config_name" : "est-rus",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 613,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 52902
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 52902,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "fra-ile" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 394,
+      "config_name" : "fra-ile",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 394,
+            "name" : "test",
+            "num_bytes" : 26879
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 26879,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "deu-tgl" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 326,
+      "config_name" : "deu-tgl",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 326,
+            "num_bytes" : 28456
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 28456,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "spa-tat" : {
+      "size_in_bytes" : 30965,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 441,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 30965
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 441,
+      "config_name" : "spa-tat",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "deu-epo" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 1429646,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 13439,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 1429646,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 13439
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2859292,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 26878,
+      "config_name" : "deu-epo"
+   },
+   "epo-hun" : {
+      "splits" : {
+         "test" : {
+            "num_examples" : 10094,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 843261
+         },
+         "validation" : {
+            "num_bytes" : 843261,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10094,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1686522,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "epo-hun",
+      "dataset_size" : 20188
+   },
+   "epo-hrv" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 846,
+      "config_name" : "epo-hrv",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 33262,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 423
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 423,
+            "num_bytes" : 33262
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 66524
+   },
+   "vie-zho" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 45707,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 440,
+            "name" : "test",
+            "num_bytes" : 45707
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 440,
+      "config_name" : "vie-zho",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "dan-deu" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 19996,
+      "config_name" : "dan-deu",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 9998,
+            "num_bytes" : 819673
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 9998,
+            "num_bytes" : 819673
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1639346
+   },
+   "eng-tur" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-tur",
+      "dataset_size" : 27814,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2267380,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 13907,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1133690
+         },
+         "test" : {
+            "num_bytes" : 1133690,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 13907
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "ell-epo" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 58730,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 58730,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 604
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 604,
+      "config_name" : "ell-epo",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "deu-dsb" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 640,
+      "config_name" : "deu-dsb",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 640,
+            "num_bytes" : 48185
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 48185
+   },
+   "jpn_Kana-rus" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 75864,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 300,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 37932
+         },
+         "validation" : {
+            "num_bytes" : 37932,
+            "num_examples" : 300,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "jpn_Kana-rus",
+      "dataset_size" : 600,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-eng" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 1124615,
+            "num_examples" : 12062,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 1124615,
+            "name" : "validation",
+            "num_examples" : 12062,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 2249230,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-eng",
+      "dataset_size" : 24124,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "deu-kur_Latn" : {
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 14913,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 223,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 14913,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "deu-kur_Latn",
+      "dataset_size" : 223,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "deu-hrx" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 32127,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 471
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 32127,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "deu-hrx",
+      "dataset_size" : 471
+   },
+   "eng-khm" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 726,
+      "config_name" : "eng-khm",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 726,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 71191
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 71191,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "swe-tur" : {
+      "dataset_size" : 203,
+      "config_name" : "swe-tur",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 14122,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 14122,
+            "num_examples" : 203,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      }
+   },
+   "heb-ukr" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "heb-ukr",
+      "dataset_size" : 966,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 83517,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 83517,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 966,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "jpn-rus" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 10142,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1259417
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10142,
+            "name" : "validation",
+            "num_bytes" : 1259417
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2518834,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 20284,
+      "config_name" : "jpn-rus",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "fra-rus" : {
+      "size_in_bytes" : 2147150,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 11490,
+            "num_bytes" : 1073575
+         },
+         "validation" : {
+            "num_bytes" : 1073575,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 11490
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 22980,
+      "config_name" : "fra-rus",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "grn-spa" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1318,
+      "config_name" : "grn-spa",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 45645,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 659
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 659,
+            "num_bytes" : 45645
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 91290,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "orv-ukr" : {
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 95946,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 95946,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 973
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "orv-ukr",
+      "dataset_size" : 973,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "isl-ita" : {
+      "size_in_bytes" : 16140,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 16140,
+            "name" : "test",
+            "num_examples" : 236,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 236,
+      "config_name" : "isl-ita",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "ces-deu" : {
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 565074,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 282537,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 3490,
+            "name" : "test"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 3490,
+            "name" : "validation",
+            "num_bytes" : 282537
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "ces-deu",
+      "dataset_size" : 6980,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "hbs-rus" : {
+      "splits" : {
+         "test" : {
+            "num_examples" : 2500,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 237071
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2500,
+            "num_bytes" : 237071
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 474142,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 5000,
+      "config_name" : "hbs-rus"
+   },
+   "deu-cmn_Hant" : {
+      "dataset_size" : 2484,
+      "config_name" : "deu-cmn_Hant",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "size_in_bytes" : 186312,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 93156,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1242
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1242,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 93156
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "eng-mar" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2278206,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 1139103,
+            "num_examples" : 10396,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_examples" : 10396,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1139103
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-mar",
+      "dataset_size" : 20792,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "egl-ita" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 10674,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 202,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 10674
+         }
+      },
+      "config_name" : "egl-ita",
+      "dataset_size" : 202,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "epo-fin" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 413316,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 206658,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2862
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2862,
+            "num_bytes" : 206658
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 5724,
+      "config_name" : "epo-fin",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fra-oci" : {
+      "dataset_size" : 806,
+      "config_name" : "fra-oci",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 68132,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 806,
+            "name" : "test",
+            "num_bytes" : 68132
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "deu-ron" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 169208,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1141,
+            "num_bytes" : 84604
+         },
+         "validation" : {
+            "num_bytes" : 84604,
+            "name" : "validation",
+            "num_examples" : 1141,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "deu-ron",
+      "dataset_size" : 2282,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "lfn_Latn-por" : {
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 504452,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 1725,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 252226
+         },
+         "validation" : {
+            "num_bytes" : 252226,
+            "name" : "validation",
+            "num_examples" : 1725,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 3450,
+      "config_name" : "lfn_Latn-por",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "uig_Arab-cmn_Hant" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 55221,
+            "name" : "test",
+            "num_examples" : 542,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 55221,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "uig_Arab-cmn_Hant",
+      "dataset_size" : 542,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "bel-rus" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 730044,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 365022,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "validation"
+         },
+         "test" : {
+            "num_bytes" : 365022,
+            "name" : "test",
+            "num_examples" : 2500,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 5000,
+      "config_name" : "bel-rus",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fin-pol" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 609,
+      "config_name" : "fin-pol",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 609,
+            "name" : "test",
+            "num_bytes" : 43252
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 43252,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "deu-kor" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1104,
+      "config_name" : "deu-kor",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 110132,
+            "num_examples" : 1104,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 110132,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "epo-jpn_Hira" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 289591,
+            "name" : "test",
+            "num_examples" : 2986,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2986,
+            "num_bytes" : 289591
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 579182,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 5972,
+      "config_name" : "epo-jpn_Hira"
+   },
+   "hun-ukr" : {
+      "config_name" : "hun-ukr",
+      "dataset_size" : 473,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "size_in_bytes" : 37775,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 473,
+            "num_bytes" : 37775
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "fin-jpn_Hani" : {
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 148968,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 976,
+            "num_bytes" : 74484
+         },
+         "test" : {
+            "num_bytes" : 74484,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 976
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 1952,
+      "config_name" : "fin-jpn_Hani",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "fra-tlh" : {
+      "dataset_size" : 620,
+      "config_name" : "fra-tlh",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 41550,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 41550,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 620,
+            "name" : "test"
+         }
+      }
+   },
+   "jpn-tlh" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jpn-tlh",
+      "dataset_size" : 674,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 65852,
+            "num_examples" : 674,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 65852,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "hbs-jpn" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 35763,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 400,
+            "name" : "test",
+            "num_bytes" : 35763
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 400,
+      "config_name" : "hbs-jpn",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "epo-nld" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 12841,
+            "num_bytes" : 1024905
+         },
+         "validation" : {
+            "num_examples" : 12841,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1024905
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 2049810,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "epo-nld",
+      "dataset_size" : 25682
+   },
+   "fin-swe" : {
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 2841,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 209037
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2841,
+            "name" : "test",
+            "num_bytes" : 209037
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 418074,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "fin-swe",
+      "dataset_size" : 5682,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ron-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ron-spa",
+      "dataset_size" : 1959,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 137180,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 137180,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1959,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "rus-zho" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 520242,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 2500,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 260121
+         },
+         "test" : {
+            "num_bytes" : 260121,
+            "name" : "test",
+            "num_examples" : 2500,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 5000,
+      "config_name" : "rus-zho",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "jpn-jpn" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "jpn-jpn",
+      "dataset_size" : 584,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 584,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 74694
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 74694
+   },
+   "fra-jpn" : {
+      "config_name" : "fra-jpn",
+      "dataset_size" : 20338,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2205628,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 1102814,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10169
+         },
+         "test" : {
+            "num_bytes" : 1102814,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10169,
+            "name" : "test"
+         }
+      }
+   },
+   "jpn_Hira-swe" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 70646,
+            "name" : "test",
+            "num_examples" : 729,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 70646,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jpn_Hira-swe",
+      "dataset_size" : 729,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ara-jpn_Hira" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ara-jpn_Hira",
+      "dataset_size" : 1092,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1092,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 111206
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 111206,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "bre-fra" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 4988,
+      "config_name" : "bre-fra",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 2494,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 145228
+         },
+         "validation" : {
+            "num_examples" : 2494,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 145228
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 290456,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "eng-tha" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1154,
+      "config_name" : "eng-tha",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 1154,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 120449
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 120449,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "ita-vie" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ita-vie",
+      "dataset_size" : 250,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 21599,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 250,
+            "num_bytes" : 21599
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      }
+   },
+   "ina-tlh" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 12352,
+            "num_examples" : 216,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 12352,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 216,
+            "name" : "validation"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 24704,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 432,
+      "config_name" : "ina-tlh"
+   },
+   "hbs-pol" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 33128,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 417,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 33128
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 417,
+      "config_name" : "hbs-pol",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "jpn-lit" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jpn-lit",
+      "dataset_size" : 246,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 21568,
+            "num_examples" : 246,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 21568,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "nds-rus" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1850,
+      "config_name" : "nds-rus",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 157244,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 925,
+            "name" : "test",
+            "num_bytes" : 78622
+         },
+         "validation" : {
+            "num_bytes" : 78622,
+            "num_examples" : 925,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "eng-ina" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 947844,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 473922,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 4988
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 4988,
+            "name" : "test",
+            "num_bytes" : 473922
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "eng-ina",
+      "dataset_size" : 9976,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "jpn-pol" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 19998,
+      "config_name" : "jpn-pol",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2089750,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 9999,
+            "num_bytes" : 1044875
+         },
+         "test" : {
+            "num_bytes" : 1044875,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 9999
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "fra-pol" : {
+      "dataset_size" : 6174,
+      "config_name" : "fra-pol",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 512254,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 256127,
+            "num_examples" : 3087,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 3087,
+            "name" : "validation",
+            "num_bytes" : 256127
+         }
+      }
+   },
+   "epo-lat" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "epo-lat",
+      "dataset_size" : 5600,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 791104,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2800,
+            "name" : "test",
+            "num_bytes" : 395552
+         },
+         "validation" : {
+            "num_examples" : 2800,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 395552
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "jpn_Hira-cmn_Hans" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 2386,
+      "config_name" : "jpn_Hira-cmn_Hans",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 251374,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1193,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 125687
+         },
+         "test" : {
+            "num_bytes" : 125687,
+            "num_examples" : 1193,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "zsm_Latn-ind" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "zsm_Latn-ind",
+      "dataset_size" : 225,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 23540,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 225,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 23540
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "fra-jpn_Hani" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "fra-jpn_Hani",
+      "dataset_size" : 2358,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 116124,
+            "name" : "test",
+            "num_examples" : 1179,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_examples" : 1179,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 116124
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 232248,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "eng-msa" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 392371,
+            "num_examples" : 5000,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 5000,
+            "name" : "validation",
+            "num_bytes" : 392371
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 784742,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-msa",
+      "dataset_size" : 10000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-ita" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 34640,
+      "config_name" : "eng-ita",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 17320,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1233733
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 17320,
+            "name" : "validation",
+            "num_bytes" : 1233733
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 2467466,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "ces-spa" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1807,
+            "name" : "validation",
+            "num_bytes" : 136343
+         },
+         "test" : {
+            "num_bytes" : 136343,
+            "num_examples" : 1807,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 272686,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "ces-spa",
+      "dataset_size" : 3614
+   },
+   "nld-tur" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "size_in_bytes" : 368554,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2500,
+            "num_bytes" : 184277
+         },
+         "test" : {
+            "num_bytes" : 184277,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "nld-tur",
+      "dataset_size" : 5000,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "nob-nno" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 31873,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 466,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 31873
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "nob-nno",
+      "dataset_size" : 466,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "dan-spa" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 5000,
+            "num_bytes" : 378035
+         },
+         "test" : {
+            "num_bytes" : 378035,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 5000
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 756070,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "dan-spa",
+      "dataset_size" : 10000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ber-epo" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ber-epo",
+      "dataset_size" : 1576,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 133156,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1576,
+            "num_bytes" : 133156
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "glg-por" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "glg-por",
+      "dataset_size" : 433,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 433,
+            "name" : "test",
+            "num_bytes" : 33423
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 33423,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "fra-swe" : {
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_examples" : 1407,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 101365
+         },
+         "validation" : {
+            "num_examples" : 1407,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 101365
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 202730,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "fra-swe",
+      "dataset_size" : 2814,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "jpn-swe" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 896,
+      "config_name" : "jpn-swe",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 82973,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 82973,
+            "name" : "test",
+            "num_examples" : 896,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "jpn_Hira-tur" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 71606,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 732,
+            "name" : "test",
+            "num_bytes" : 71606
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "jpn_Hira-tur",
+      "dataset_size" : 732,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "ber_Latn-deu" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 55766,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 659,
+            "name" : "test",
+            "num_bytes" : 55766
+         }
+      },
+      "config_name" : "ber_Latn-deu",
+      "dataset_size" : 659,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "heb-heb" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 192758,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 96379,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1000
+         },
+         "test" : {
+            "num_bytes" : 96379,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1000
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "heb-heb",
+      "dataset_size" : 2000,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fin-jpn" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fin-jpn",
+      "dataset_size" : 19868,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1877160,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 938580,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 9934
+         },
+         "validation" : {
+            "num_bytes" : 938580,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 9934,
+            "name" : "validation"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "fra-toki" : {
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 48138,
+            "num_examples" : 562,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_examples" : 562,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 48138
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 96276,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1124,
+      "config_name" : "fra-toki",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "jpn-toki" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "jpn-toki",
+      "dataset_size" : 242,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 20860,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 242,
+            "name" : "test",
+            "num_bytes" : 20860
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "eng-kor_Hang" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 413158,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2376,
+            "num_bytes" : 206579
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 2376,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 206579
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 4752,
+      "config_name" : "eng-kor_Hang",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "tur-cmn_Hans" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 40855,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 585,
+            "name" : "test",
+            "num_bytes" : 40855
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "tur-cmn_Hans",
+      "dataset_size" : 585,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fra-lfn_Latn" : {
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 65170,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 361,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 32585
+         },
+         "test" : {
+            "num_bytes" : 32585,
+            "num_examples" : 361,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "fra-lfn_Latn",
+      "dataset_size" : 722,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "ina-lad" : {
+      "size_in_bytes" : 46940,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 320,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 23470
+         },
+         "test" : {
+            "num_bytes" : 23470,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 320
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ina-lad",
+      "dataset_size" : 640,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "ita-tur" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ita-tur",
+      "dataset_size" : 20000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10000,
+            "name" : "test",
+            "num_bytes" : 899367
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10000,
+            "name" : "validation",
+            "num_bytes" : 899367
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1798734,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "bul-ita" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500,
+            "num_bytes" : 241262
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "validation",
+            "num_bytes" : 241262
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 482524,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "bul-ita",
+      "dataset_size" : 5000
+   },
+   "eng-swa" : {
+      "config_name" : "eng-swa",
+      "dataset_size" : 387,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "size_in_bytes" : 22845,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 22845,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 387,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "fin-kur" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 259,
+      "config_name" : "fin-kur",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 20711,
+            "num_examples" : 259,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 20711,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "eng-ota" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 58713,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 58713,
+            "num_examples" : 688,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-ota",
+      "dataset_size" : 688,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "por-tur" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "por-tur",
+      "dataset_size" : 1794,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 125270,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1794
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 125270,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "eng-srp_Latn" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 994016,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 497008,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 6656
+         },
+         "validation" : {
+            "num_examples" : 6656,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 497008
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 13312,
+      "config_name" : "eng-srp_Latn",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fra-toki_Latn" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 48117,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 561,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 48117,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 561,
+      "config_name" : "fra-toki_Latn",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "bel-fra" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 283,
+      "config_name" : "bel-fra",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 283,
+            "num_bytes" : 26819
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 26819
+   },
+   "heb-jpn_Hira" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 201,
+      "config_name" : "heb-jpn_Hira",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 201,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 23573
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 23573,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "deu-ido" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1742,
+      "config_name" : "deu-ido",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 58820,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 871,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 58820,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 871
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 117640,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "rus-cmn_Hant" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 149490,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 74745,
+            "name" : "test",
+            "num_examples" : 799,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 799,
+            "name" : "validation",
+            "num_bytes" : 74745
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "rus-cmn_Hant",
+      "dataset_size" : 1598,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "slv-zho" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 825,
+      "config_name" : "slv-zho",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 60602,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 825,
+            "num_bytes" : 60602
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "heb-nld" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2500,
+            "num_bytes" : 236821
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500,
+            "num_bytes" : 236821
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 473642,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "heb-nld",
+      "dataset_size" : 5000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "lat-pol" : {
+      "dataset_size" : 915,
+      "config_name" : "lat-pol",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 61028,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 61028,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 915,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null
+   },
+   "hun-hun" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1064,
+            "name" : "test",
+            "num_bytes" : 84532
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 84532,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "hun-hun",
+      "dataset_size" : 1064
+   },
+   "gcf-gcf" : {
+      "size_in_bytes" : 14350,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_examples" : 233,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 14350
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "gcf-gcf",
+      "dataset_size" : 233,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "cat-ita" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 298,
+      "config_name" : "cat-ita",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 22469,
+            "num_examples" : 298,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 22469,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "fra-run" : {
+      "size_in_bytes" : 72697,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1274,
+            "num_bytes" : 72697
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1274,
+      "config_name" : "fra-run",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "chv-eng" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "chv-eng",
+      "dataset_size" : 336,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 28676,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 28676,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 336,
+            "name" : "test"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "lit-tur" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1472,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 102387
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 102387,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1472,
+      "config_name" : "lit-tur"
+   },
+   "afr-nld" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 68060,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 68060,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1056
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "afr-nld",
+      "dataset_size" : 1056,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-slv" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 4990,
+      "config_name" : "eng-slv",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2495,
+            "name" : "test",
+            "num_bytes" : 166748
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2495,
+            "name" : "validation",
+            "num_bytes" : 166748
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 333496,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "ber-deu" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 661,
+      "config_name" : "ber-deu",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 661,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 56019
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 56019,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "deu-spa" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 21042,
+      "config_name" : "deu-spa",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 10521,
+            "num_bytes" : 930791
+         },
+         "validation" : {
+            "num_bytes" : 930791,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10521,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1861582,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "fas-fra" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 376,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 45471
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 45471,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "fas-fra",
+      "dataset_size" : 376
+   },
+   "nor-spa" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 79106,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 79106,
+            "name" : "test",
+            "num_examples" : 960,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 960,
+      "config_name" : "nor-spa",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-got" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 202,
+      "config_name" : "eng-got",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 202,
+            "num_bytes" : 20327
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 20327
+   },
+   "heb-lat" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 23651,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 271
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 271,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 23651
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 47302,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "heb-lat",
+      "dataset_size" : 542
+   },
+   "spa-ukr" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1635046,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 817523,
+            "name" : "validation",
+            "num_examples" : 10115,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 817523,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10115,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "spa-ukr",
+      "dataset_size" : 20230,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "ell-spa" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ell-spa",
+      "dataset_size" : 3658,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 1829,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 155748
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1829,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 155748
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 311496,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "chv-tur" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 508,
+      "config_name" : "chv-tur",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 44708,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 508,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 44708,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "ara-heb" : {
+      "config_name" : "ara-heb",
+      "dataset_size" : 2416,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 221514,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 110757,
+            "num_examples" : 1208,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1208,
+            "name" : "validation",
+            "num_bytes" : 110757
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "eng-mal" : {
+      "dataset_size" : 802,
+      "config_name" : "eng-mal",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "size_in_bytes" : 113410,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 113410,
+            "num_examples" : 802,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "rus-xal" : {
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 23410,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 23410,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 209
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "rus-xal",
+      "dataset_size" : 209,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "dtp-eng" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 1927,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 141337
+         },
+         "test" : {
+            "num_bytes" : 141337,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1927,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 282674,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "dtp-eng",
+      "dataset_size" : 3854,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "fra-jbo" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1114,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 94434
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 94434,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1114,
+      "config_name" : "fra-jbo"
+   },
+   "eng-uig" : {
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 314767,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 3024
+         },
+         "test" : {
+            "num_bytes" : 314767,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 3024,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 629534,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 6048,
+      "config_name" : "eng-uig"
+   },
+   "eng-nob" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 9078,
+      "config_name" : "eng-nob",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 370388,
+            "name" : "validation",
+            "num_examples" : 4539,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 4539,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 370388
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 740776
+   },
+   "deu-yid" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 79442,
+            "name" : "validation",
+            "num_examples" : 853,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 853,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 79442
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 158884,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1706,
+      "config_name" : "deu-yid"
+   },
+   "srp_Latn-ita" : {
+      "size_in_bytes" : 14324,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 212,
+            "name" : "test",
+            "num_bytes" : 14324
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "srp_Latn-ita",
+      "dataset_size" : 212,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "nds-por" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 12995,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 12995,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 207
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 207,
+      "config_name" : "nds-por",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "ind-zsm_Latn" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 225,
+            "num_bytes" : 23472
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 23472,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 225,
+      "config_name" : "ind-zsm_Latn"
+   },
+   "deu-frr" : {
+      "config_name" : "deu-frr",
+      "dataset_size" : 278,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 19409,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 19409,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 278
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "eng-nst" : {
+      "splits" : {
+         "test" : {
+            "num_bytes" : 61426,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 805,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 61426,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "eng-nst",
+      "dataset_size" : 805
+   },
+   "lat-tlh" : {
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 28222,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 14111,
+            "name" : "validation",
+            "num_examples" : 233,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 14111,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 233
+         }
+      },
+      "dataset_size" : 466,
+      "config_name" : "lat-tlh",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "epo-srp_Cyrl" : {
+      "dataset_size" : 3590,
+      "config_name" : "epo-srp_Cyrl",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "size_in_bytes" : 407002,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 203501,
+            "num_examples" : 1795,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_examples" : 1795,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 203501
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "epo-heb" : {
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 1976134,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 988067,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10368
+         },
+         "test" : {
+            "num_bytes" : 988067,
+            "num_examples" : 10368,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "epo-heb",
+      "dataset_size" : 20736,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "bul-eng" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 20000,
+      "config_name" : "bul-eng",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 1928748,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 964374,
+            "name" : "validation",
+            "num_examples" : 10000,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_examples" : 10000,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 964374
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "fin-por" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 31930,
+            "name" : "test",
+            "num_examples" : 477,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 31930,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "fin-por",
+      "dataset_size" : 477,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-vie" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 5000,
+      "config_name" : "eng-vie",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500,
+            "num_bytes" : 234244
+         },
+         "validation" : {
+            "num_examples" : 2500,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 234244
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 468488
+   },
+   "lat-rus" : {
+      "size_in_bytes" : 191186,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 95593,
+            "name" : "validation",
+            "num_examples" : 1041,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 95593,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1041
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 2082,
+      "config_name" : "lat-rus",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "ita-ita" : {
+      "config_name" : "ita-ita",
+      "dataset_size" : 2000,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 159922,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1000,
+            "name" : "validation",
+            "num_bytes" : 79961
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1000,
+            "name" : "test",
+            "num_bytes" : 79961
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "ita-msa" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 32292,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 426
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 32292,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 426,
+      "config_name" : "ita-msa",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "bul-tur" : {
+      "size_in_bytes" : 88651,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 834,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 88651
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 834,
+      "config_name" : "bul-tur",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "fra-cmn_Hans" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 13650,
+      "config_name" : "fra-cmn_Hans",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 632952,
+            "name" : "validation",
+            "num_examples" : 6825,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 632952,
+            "num_examples" : 6825,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1265904,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "ara-ber_Latn" : {
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 2299,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 208463
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2299,
+            "name" : "test",
+            "num_bytes" : 208463
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 416926,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ara-ber_Latn",
+      "dataset_size" : 4598
+   },
+   "ina-por" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 551368,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 275684,
+            "name" : "validation",
+            "num_examples" : 2537,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2537,
+            "num_bytes" : 275684
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 5074,
+      "config_name" : "ina-por",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "hun-lat" : {
+      "config_name" : "hun-lat",
+      "dataset_size" : 607,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 43106,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 43106,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 607,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "fin-fra" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 263504,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 1920,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 131752
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1920,
+            "name" : "validation",
+            "num_bytes" : 131752
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "fin-fra",
+      "dataset_size" : 3840,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "fra-ind" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 75212,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 905,
+            "name" : "test",
+            "num_bytes" : 75212
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 905,
+      "config_name" : "fra-ind",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-yue_Hant" : {
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 70248,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 35124,
+            "num_examples" : 454,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 35124,
+            "num_examples" : 454,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 908,
+      "config_name" : "eng-yue_Hant",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "wuu-cmn_Hans" : {
+      "config_name" : "wuu-cmn_Hans",
+      "dataset_size" : 1746,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 154860,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 77430,
+            "num_examples" : 873,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 77430,
+            "name" : "validation",
+            "num_examples" : 873,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null
+   },
+   "ceb-deu" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ceb-deu",
+      "dataset_size" : 903,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 76229,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 903,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 76229
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "cat-eng" : {
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1631,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 126706
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1631,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 126706
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 253412,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 3262,
+      "config_name" : "cat-eng"
+   },
+   "epo-nds" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 183770,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2536
+         },
+         "validation" : {
+            "num_bytes" : 183770,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2536
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 367540,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "epo-nds",
+      "dataset_size" : 5072,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "heb-hun" : {
+      "dataset_size" : 417,
+      "config_name" : "heb-hun",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 32411,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 417,
+            "num_bytes" : 32411
+         }
+      },
+      "task_templates" : null
+   },
+   "jpn-por" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 386608,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 193304,
+            "num_examples" : 1940,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1940,
+            "num_bytes" : 193304
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 3880,
+      "config_name" : "jpn-por",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fra-por" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 21036,
+      "config_name" : "fra-por",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 1716664,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 858332,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10518,
+            "name" : "validation"
+         },
+         "test" : {
+            "num_bytes" : 858332,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10518,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "ces-epo" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 525606,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 3471,
+            "num_bytes" : 262803
+         },
+         "validation" : {
+            "num_examples" : 3471,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 262803
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 6942,
+      "config_name" : "ces-epo",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "rus-sah" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 117992,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 117992,
+            "name" : "test",
+            "num_examples" : 994,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "config_name" : "rus-sah",
+      "dataset_size" : 994,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "eng-vol" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1541,
+            "num_bytes" : 100559
+         },
+         "test" : {
+            "num_bytes" : 100559,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1541,
+            "name" : "test"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 201118,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-vol",
+      "dataset_size" : 3082
+   },
+   "eng-hbs" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-hbs",
+      "dataset_size" : 20034,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 1514106,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 757053,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10017
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10017,
+            "name" : "test",
+            "num_bytes" : 757053
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "hun-nld" : {
+      "size_in_bytes" : 227698,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 113849,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1629
+         },
+         "test" : {
+            "num_bytes" : 113849,
+            "name" : "test",
+            "num_examples" : 1629,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 3258,
+      "config_name" : "hun-nld",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "dtp-msa" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 517,
+      "config_name" : "dtp-msa",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 53688,
+            "name" : "test",
+            "num_examples" : 517,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 53688,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null
+   },
+   "epo-ukr" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 5000,
+      "config_name" : "epo-ukr",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 403978,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 2500,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 201989
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 2500,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 201989
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "hun-jpn_Hira" : {
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 242531,
+            "name" : "test",
+            "num_examples" : 2127,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_examples" : 2127,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 242531
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 485062,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 4254,
+      "config_name" : "hun-jpn_Hira",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "hrv-pol" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "size_in_bytes" : 19473,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 271,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 19473
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 271,
+      "config_name" : "hrv-pol",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "fra-fra" : {
+      "config_name" : "fra-fra",
+      "dataset_size" : 2000,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 168264,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1000,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 84132
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1000,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 84132
+         }
+      }
+   },
+   "fra-kab" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 24982,
+      "config_name" : "fra-kab",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 2103388,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 1051694,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 12491,
+            "name" : "validation"
+         },
+         "test" : {
+            "num_bytes" : 1051694,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 12491
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "deu-lfn" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 86654,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 418,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 43327
+         },
+         "validation" : {
+            "num_bytes" : 43327,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 418
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 836,
+      "config_name" : "deu-lfn",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "deu-zho" : {
+      "dataset_size" : 7888,
+      "config_name" : "deu-zho",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 654652,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 327326,
+            "name" : "test",
+            "num_examples" : 3944,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 327326,
+            "num_examples" : 3944,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "dan-epo" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "dan-epo",
+      "dataset_size" : 22138,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1686702,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 843351,
+            "name" : "validation",
+            "num_examples" : 11069,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_examples" : 11069,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 843351
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "deu-deu" : {
+      "dataset_size" : 5000,
+      "config_name" : "deu-deu",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 476026,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 238013,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 238013,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2500
+         }
+      }
+   },
+   "nor-zho" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 201,
+      "config_name" : "nor-zho",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 201,
+            "name" : "test",
+            "num_bytes" : 16243
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 16243,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "deu-isl" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "deu-isl",
+      "dataset_size" : 969,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 68421,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 68421,
+            "name" : "test",
+            "num_examples" : 969,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "ber-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 20032,
+      "config_name" : "ber-spa",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 1615252,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 10016,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 807626
+         },
+         "validation" : {
+            "num_bytes" : 807626,
+            "num_examples" : 10016,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "pol-rus" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 3543,
+            "name" : "validation",
+            "num_bytes" : 333617
+         },
+         "test" : {
+            "num_examples" : 3543,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 333617
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 667234,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 7086,
+      "config_name" : "pol-rus"
+   },
+   "eng-hin" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-hin",
+      "dataset_size" : 10000,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1144976,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 572488,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 5000
+         },
+         "validation" : {
+            "num_examples" : 5000,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 572488
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "por-tgl" : {
+      "size_in_bytes" : 152019,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 152019,
+            "name" : "test",
+            "num_examples" : 1777,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1777,
+      "config_name" : "por-tgl",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "fra-nds" : {
+      "size_in_bytes" : 59528,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 857,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 59528
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fra-nds",
+      "dataset_size" : 857,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "jpn-nds" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 69132,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 406,
+            "num_bytes" : 34566
+         },
+         "test" : {
+            "num_bytes" : 34566,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 406,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 812,
+      "config_name" : "jpn-nds",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "lfn-yid" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 187234,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 93617,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 993
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 993,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 93617
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "lfn-yid",
+      "dataset_size" : 1986,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "dan-ita" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "dan-ita",
+      "dataset_size" : 284,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 23819,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 284,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 23819
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "deu-nob" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 3525,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 363529
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 3525,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 363529
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 727058,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "deu-nob",
+      "dataset_size" : 7050
+   },
+   "eng-yid" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 4966,
+      "config_name" : "eng-yid",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 464316,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 232158,
+            "num_examples" : 2483,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 232158,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2483,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "hun-jpn_Hani" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 351,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 37322
+         },
+         "validation" : {
+            "num_examples" : 351,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 37322
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 74644,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "hun-jpn_Hani",
+      "dataset_size" : 702,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "hrv-jpn_Hira" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 24963,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 276,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 24963
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 276,
+      "config_name" : "hrv-jpn_Hira",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "dsb-slv" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 330,
+      "config_name" : "dsb-slv",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 330,
+            "num_bytes" : 16248
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 16248,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "lad_Latn-lfn_Latn" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 422,
+      "config_name" : "lad_Latn-lfn_Latn",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 13412,
+            "name" : "validation",
+            "num_examples" : 211,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 211,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 13412
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 26824,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "hun-pol" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1934,
+      "config_name" : "hun-pol",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1934,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 140082
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 140082,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "hun-swe" : {
+      "size_in_bytes" : 235874,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1614,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 117937
+         },
+         "test" : {
+            "num_bytes" : 117937,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1614
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "hun-swe",
+      "dataset_size" : 3228,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "epo-cmn_Hans" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "epo-cmn_Hans",
+      "dataset_size" : 2808,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 111383,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1404,
+            "name" : "test"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1404,
+            "name" : "validation",
+            "num_bytes" : 111383
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 222766
+   },
+   "eng-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 33166,
+      "config_name" : "eng-spa",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2841844,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 1420922,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 16583
+         },
+         "test" : {
+            "num_bytes" : 1420922,
+            "num_examples" : 16583,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "ces-ita" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1099,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 85180
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 85180,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ces-ita",
+      "dataset_size" : 1099
+   },
+   "lfn-spa" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 21163,
+            "num_examples" : 264,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 21163,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 264,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 42326,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 528,
+      "config_name" : "lfn-spa",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "lat_Latn-lfn_Latn" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "lat_Latn-lfn_Latn",
+      "dataset_size" : 598,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 38204,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 299,
+            "num_bytes" : 19102
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 299,
+            "num_bytes" : 19102
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "deu-slv" : {
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 492,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 30184
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 30184,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 492,
+      "config_name" : "deu-slv",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ben-eng" : {
+      "dataset_size" : 5000,
+      "config_name" : "ben-eng",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 469172,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 234586,
+            "num_examples" : 2500,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500,
+            "num_bytes" : 234586
+         }
+      }
+   },
+   "nld-zho" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "nld-zho",
+      "dataset_size" : 3308,
+      "splits" : {
+         "test" : {
+            "num_examples" : 1654,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 140786
+         },
+         "validation" : {
+            "num_bytes" : 140786,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1654,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 281572,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "afr-rus" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 228,
+      "config_name" : "afr-rus",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 17527,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 17527,
+            "name" : "test",
+            "num_examples" : 228,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "fin-jpn_Kana" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 452,
+      "config_name" : "fin-jpn_Kana",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 20953,
+            "num_examples" : 226,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 20953,
+            "name" : "validation",
+            "num_examples" : 226,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 41906,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "awa-eng" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 20696,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 279,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 20696,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "awa-eng",
+      "dataset_size" : 279
+   },
+   "eng-ido" : {
+      "dataset_size" : 3920,
+      "config_name" : "eng-ido",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 298258,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1960,
+            "name" : "validation",
+            "num_bytes" : 149129
+         },
+         "test" : {
+            "num_bytes" : 149129,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1960
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "jpn_Hira-nds" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jpn_Hira-nds",
+      "dataset_size" : 694,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 29378,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 347
+         },
+         "validation" : {
+            "num_bytes" : 29378,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 347,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 58756,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "por-cmn_Hant" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 26065,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 386
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 26065,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "por-cmn_Hant",
+      "dataset_size" : 386,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "hbs-ukr" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 76362,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 76362,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 942
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "hbs-ukr",
+      "dataset_size" : 942,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "run-rus" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "run-rus",
+      "dataset_size" : 1251,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1251,
+            "num_bytes" : 86691
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 86691
+   },
+   "uig-zho" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1929,
+      "config_name" : "uig-zho",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1929,
+            "num_bytes" : 207018
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 207018
+   },
+   "eng-kaz" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 403,
+      "config_name" : "eng-kaz",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 37502,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 403,
+            "num_bytes" : 37502
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "ita-cmn_Hant" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_examples" : 1037,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 72270
+         },
+         "validation" : {
+            "num_bytes" : 72270,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1037,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 144540,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 2074,
+      "config_name" : "ita-cmn_Hant",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eus-rus" : {
+      "size_in_bytes" : 35444,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 485,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 35444
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eus-rus",
+      "dataset_size" : 485,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "eng-kat" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-kat",
+      "dataset_size" : 984,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 94570,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 984,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 94570
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "jpn-ukr" : {
+      "dataset_size" : 914,
+      "config_name" : "jpn-ukr",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 97488,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 457,
+            "name" : "test",
+            "num_bytes" : 48744
+         },
+         "validation" : {
+            "num_examples" : 457,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 48744
+         }
+      }
+   },
+   "fra-ukr" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fra-ukr",
+      "dataset_size" : 20070,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1672946,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 836473,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10035,
+            "name" : "validation"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 10035,
+            "num_bytes" : 836473
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "heb-rus" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 5000,
+      "config_name" : "heb-rus",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "validation",
+            "num_bytes" : 272864
+         },
+         "test" : {
+            "num_bytes" : 272864,
+            "name" : "test",
+            "num_examples" : 2500,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 545728,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "epo-fra" : {
+      "dataset_size" : 24336,
+      "config_name" : "epo-fra",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "size_in_bytes" : 2168862,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 12168,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1084431
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 12168,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1084431
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "jbo_Latn-cmn_Hans" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 20111,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 281,
+            "num_bytes" : 20111
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "jbo_Latn-cmn_Hans",
+      "dataset_size" : 281,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "lfn_Cyrl-por" : {
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 56978,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 28489,
+            "num_examples" : 220,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 220,
+            "name" : "validation",
+            "num_bytes" : 28489
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "lfn_Cyrl-por",
+      "dataset_size" : 440,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "lat-nor" : {
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_examples" : 333,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 23606
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 23606,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "lat-nor",
+      "dataset_size" : 333,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "deu-srp_Latn" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 94067,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 986,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 94067
+         }
+      },
+      "dataset_size" : 986,
+      "config_name" : "deu-srp_Latn",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "epo-por" : {
+      "config_name" : "epo-por",
+      "dataset_size" : 21618,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "size_in_bytes" : 2077082,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 10809,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1038541
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10809,
+            "name" : "validation",
+            "num_bytes" : 1038541
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "rus-tur" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 10000,
+      "config_name" : "rus-tur",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 1034900,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 5000,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 517450
+         },
+         "validation" : {
+            "num_bytes" : 517450,
+            "name" : "validation",
+            "num_examples" : 5000,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "ukr-zho" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1575,
+      "config_name" : "ukr-zho",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 138693,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1575,
+            "name" : "test",
+            "num_bytes" : 138693
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "kor_Hang-rus" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 220,
+      "config_name" : "kor_Hang-rus",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 220,
+            "num_bytes" : 23000
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 23000,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "hun-jpn" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "hun-jpn",
+      "dataset_size" : 4998,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 282076,
+            "name" : "validation",
+            "num_examples" : 2499,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 282076,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2499,
+            "name" : "test"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 564152
+   },
+   "ara-ell" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 426,
+      "config_name" : "ara-ell",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 426,
+            "num_bytes" : 32889
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 32889,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "ara-fra" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "ara-fra",
+      "dataset_size" : 3138,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 134171,
+            "num_examples" : 1569,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 134171,
+            "name" : "validation",
+            "num_examples" : 1569,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 268342,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "pol-swe" : {
+      "config_name" : "pol-swe",
+      "dataset_size" : 1392,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 100023,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 100023,
+            "name" : "test",
+            "num_examples" : 1392,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "fin-heb" : {
+      "size_in_bytes" : 19871,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 212,
+            "name" : "test",
+            "num_bytes" : 19871
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fin-heb",
+      "dataset_size" : 212,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "heb-jpn" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 27471,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 241,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 27471
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "heb-jpn",
+      "dataset_size" : 241,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eus-jpn" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 14683,
+            "num_examples" : 237,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 14683,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eus-jpn",
+      "dataset_size" : 237
+   },
+   "jpn_Hira-ukr" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 383,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 41595
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 41595,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "jpn_Hira-ukr",
+      "dataset_size" : 383,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "lat-lat" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "lat-lat",
+      "dataset_size" : 235,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 20141,
+            "num_examples" : 235,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 20141,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "dan-eng" : {
+      "config_name" : "dan-eng",
+      "dataset_size" : 21590,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "size_in_bytes" : 1615962,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 10795,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 807981
+         },
+         "test" : {
+            "num_bytes" : 807981,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10795,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "heb-lad" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "heb-lad",
+      "dataset_size" : 436,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 37786,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 218,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 18893
+         },
+         "validation" : {
+            "num_examples" : 218,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 18893
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "spa-cmn_Hans" : {
+      "config_name" : "spa-cmn_Hans",
+      "dataset_size" : 2284,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 167634,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1142,
+            "name" : "test",
+            "num_bytes" : 83817
+         },
+         "validation" : {
+            "num_bytes" : 83817,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1142,
+            "name" : "validation"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "eng-lfn" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-lfn",
+      "dataset_size" : 6594,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 526450,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 3297,
+            "num_bytes" : 263225
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 3297,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 263225
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "dtp-zsm_Latn" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 52979,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 509
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 52979,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "dtp-zsm_Latn",
+      "dataset_size" : 509
+   },
+   "eng-zho" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-zho",
+      "dataset_size" : 20780,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 871802,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10390
+         },
+         "test" : {
+            "num_examples" : 10390,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 871802
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1743604
+   },
+   "nld-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 20226,
+      "config_name" : "nld-spa",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 1632458,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 10113,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 816229
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 10113,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 816229
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "eng-isl" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 5006,
+      "config_name" : "eng-isl",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 421904,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 2503,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 210952
+         },
+         "test" : {
+            "num_bytes" : 210952,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2503
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "dan-tur" : {
+      "size_in_bytes" : 44770,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 758,
+            "name" : "test",
+            "num_bytes" : 44770
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "dan-tur",
+      "dataset_size" : 758,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "isl-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "isl-spa",
+      "dataset_size" : 238,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 13806,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 13806,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 238
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "bel-ukr" : {
+      "config_name" : "bel-ukr",
+      "dataset_size" : 4710,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 569278,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2355,
+            "num_bytes" : 284639
+         },
+         "test" : {
+            "num_examples" : 2355,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 284639
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "lad-lat" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 29338,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 14669,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 214,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 14669,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 214,
+            "name" : "validation"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "lad-lat",
+      "dataset_size" : 428,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "hun-rus" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 505718,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 252859,
+            "num_examples" : 2687,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 252859,
+            "name" : "validation",
+            "num_examples" : 2687,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "hun-rus",
+      "dataset_size" : 5374,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "kor_Hang-cmn_Hans" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 207,
+      "config_name" : "kor_Hang-cmn_Hans",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 207,
+            "name" : "test",
+            "num_bytes" : 18974
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 18974
+   },
+   "jpn_Hira-tlh_Latn" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jpn_Hira-tlh_Latn",
+      "dataset_size" : 638,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 638,
+            "name" : "test",
+            "num_bytes" : 62553
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 62553,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "ita-ron" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ita-ron",
+      "dataset_size" : 1005,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1005,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 68917
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 68917,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "eng-fao" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-fao",
+      "dataset_size" : 294,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 294,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 20156
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 20156,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "kat-rus" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "kat-rus",
+      "dataset_size" : 642,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 91485,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 642,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 91485
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "cor-rus" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 14472,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 218,
+            "num_bytes" : 14472
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 218,
+      "config_name" : "cor-rus",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "swe-zho" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1049,
+      "config_name" : "swe-zho",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 68783,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1049,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 68783
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "deu-hbs" : {
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 170005,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1959,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 170005
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 1959,
+      "config_name" : "deu-hbs",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "deu-vol" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "deu-vol",
+      "dataset_size" : 206,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 206,
+            "num_bytes" : 14006
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 14006,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null
+   },
+   "ces-eng" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 27648,
+      "config_name" : "ces-eng",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 2086344,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 13824,
+            "num_bytes" : 1043172
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 13824,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1043172
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "por-ron" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1362,
+      "config_name" : "por-ron",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 98310,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 49155,
+            "num_examples" : 681,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_examples" : 681,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 49155
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "cbk-eng" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "cbk-eng",
+      "dataset_size" : 2996,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 220722,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 110361,
+            "num_examples" : 1498,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 110361,
+            "name" : "test",
+            "num_examples" : 1498,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "ido_Latn-lfn_Latn" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 618,
+      "config_name" : "ido_Latn-lfn_Latn",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 309,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 20200
+         },
+         "test" : {
+            "num_examples" : 309,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 20200
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 40400,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "cat-epo" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "cat-epo",
+      "dataset_size" : 772,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 772,
+            "num_bytes" : 61422
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 61422,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "fra-heb" : {
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 3281,
+            "name" : "validation",
+            "num_bytes" : 311865
+         },
+         "test" : {
+            "num_examples" : 3281,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 311865
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 623730,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 6562,
+      "config_name" : "fra-heb"
+   },
+   "heb-lfn_Latn" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 34598,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 212,
+            "name" : "validation",
+            "num_bytes" : 17299
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 212,
+            "num_bytes" : 17299
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "heb-lfn_Latn",
+      "dataset_size" : 424,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-lad_Latn" : {
+      "size_in_bytes" : 81274,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 40637,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 672
+         },
+         "test" : {
+            "num_bytes" : 40637,
+            "num_examples" : 672,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-lad_Latn",
+      "dataset_size" : 1344,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "eng-tuk_Latn" : {
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "size_in_bytes" : 213379,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 2499,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 213379
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "eng-tuk_Latn",
+      "dataset_size" : 2499,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "epo-jbo" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "epo-jbo",
+      "dataset_size" : 1167,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 83091,
+            "name" : "test",
+            "num_examples" : 1167,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 83091
+   },
+   "bul-bul" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 219,
+            "num_bytes" : 23925
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 23925,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 219,
+      "config_name" : "bul-bul",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ron-tur" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 407434,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 2460,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 203717
+         },
+         "validation" : {
+            "num_bytes" : 203717,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2460,
+            "name" : "validation"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 4920,
+      "config_name" : "ron-tur",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-grc" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-grc",
+      "dataset_size" : 614,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 70785,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 614,
+            "name" : "test"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 70785,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "heb-pol" : {
+      "size_in_bytes" : 874820,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 437410,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 5000
+         },
+         "test" : {
+            "num_bytes" : 437410,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 5000
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "heb-pol",
+      "dataset_size" : 10000,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "deu-vie" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "deu-vie",
+      "dataset_size" : 400,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 38820,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 38820,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 400,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "epo-fas" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 268962,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2514
+         },
+         "validation" : {
+            "num_bytes" : 268962,
+            "num_examples" : 2514,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 537924,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 5028,
+      "config_name" : "epo-fas",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "avk-fra" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 107793,
+            "name" : "test",
+            "num_examples" : 1244,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1244,
+            "name" : "validation",
+            "num_bytes" : 107793
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 215586,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "avk-fra",
+      "dataset_size" : 2488
+   },
+   "bul-epo" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 570,
+      "config_name" : "bul-epo",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 570,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 56818
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 56818,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "lat-nld" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "lat-nld",
+      "dataset_size" : 366,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 366,
+            "num_bytes" : 24902
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 24902,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "hun-por" : {
+      "dataset_size" : 5000,
+      "config_name" : "hun-por",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 352228,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 176114,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2500
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500,
+            "num_bytes" : 176114
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "eng-hrx" : {
+      "size_in_bytes" : 13943,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 13943,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 221
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 221,
+      "config_name" : "eng-hrx",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "eng-jbo_Latn" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-jbo_Latn",
+      "dataset_size" : 9984,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 4992,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 363806
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 4992,
+            "name" : "test",
+            "num_bytes" : 363806
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 727612
+   },
+   "epo-jpn" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 669924,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 334962,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 3477,
+            "name" : "validation"
+         },
+         "test" : {
+            "num_bytes" : 334962,
+            "name" : "test",
+            "num_examples" : 3477,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "epo-jpn",
+      "dataset_size" : 6954,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "eng-gsw" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-gsw",
+      "dataset_size" : 205,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 205,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 9985
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 9985,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "ina_Latn-tlh_Latn" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 215,
+            "num_bytes" : 12311
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 12311,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 215,
+      "config_name" : "ina_Latn-tlh_Latn"
+   },
+   "fra-uig_Arab" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 693,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 66689
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 66689,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "fra-uig_Arab",
+      "dataset_size" : 693,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "epo-lad" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 25590,
+            "num_examples" : 366,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 366,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 25590
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 51180,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 732,
+      "config_name" : "epo-lad",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-cmn_Hant" : {
+      "dataset_size" : 8952,
+      "config_name" : "eng-cmn_Hant",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 721472,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 360736,
+            "num_examples" : 4476,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 360736,
+            "name" : "test",
+            "num_examples" : 4476,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null
+   },
+   "jpn-nld" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 311942,
+            "num_examples" : 3430,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 311942,
+            "name" : "test",
+            "num_examples" : 3430,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 623884,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 6860,
+      "config_name" : "jpn-nld",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "fra-nld" : {
+      "size_in_bytes" : 1788096,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 894048,
+            "name" : "validation",
+            "num_examples" : 11548,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 11548,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 894048
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fra-nld",
+      "dataset_size" : 23096,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "epo-tlh" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1930,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 127748
+         },
+         "test" : {
+            "num_bytes" : 127748,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1930,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 255496,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "epo-tlh",
+      "dataset_size" : 3860,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-tel" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-tel",
+      "dataset_size" : 261,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 261,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 28967
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 28967
+   },
+   "ell-tur" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 2938,
+      "config_name" : "ell-tur",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 132670,
+            "name" : "validation",
+            "num_examples" : 1469,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_examples" : 1469,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 132670
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 265340,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "eng-tzl_Latn" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 200,
+            "name" : "test",
+            "num_bytes" : 10349
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 10349,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 200,
+      "config_name" : "eng-tzl_Latn",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-epo" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 17675,
+            "name" : "test",
+            "num_bytes" : 1455517
+         },
+         "validation" : {
+            "num_bytes" : 1455517,
+            "name" : "validation",
+            "num_examples" : 17675,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2911034,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 35350,
+      "config_name" : "eng-epo"
+   },
+   "cor-fra" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 555,
+      "config_name" : "cor-fra",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 31711,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 31711,
+            "num_examples" : 555,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "ara-jpn" : {
+      "config_name" : "ara-jpn",
+      "dataset_size" : 1336,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 136439,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_examples" : 1336,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 136439
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "eng-lfn_Cyrl" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "size_in_bytes" : 159106,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 847,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 79553
+         },
+         "validation" : {
+            "num_bytes" : 79553,
+            "name" : "validation",
+            "num_examples" : 847,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "eng-lfn_Cyrl",
+      "dataset_size" : 1694,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "deu-tur" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 10000,
+      "config_name" : "deu-tur",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 5000,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 409984
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 5000,
+            "num_bytes" : 409984
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 819968
+   },
+   "fin-lat" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "fin-lat",
+      "dataset_size" : 294,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 294,
+            "name" : "test",
+            "num_bytes" : 18696
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 18696,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "fra-jpn_Hira" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fra-jpn_Hira",
+      "dataset_size" : 17582,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 1930970,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 8791,
+            "name" : "validation",
+            "num_bytes" : 965485
+         },
+         "test" : {
+            "num_bytes" : 965485,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 8791
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "ell-eng" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 963409,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10899,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 963409,
+            "num_examples" : 10899,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1926818,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "ell-eng",
+      "dataset_size" : 21798
+   },
+   "lit-spa" : {
+      "dataset_size" : 454,
+      "config_name" : "lit-spa",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 29135,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 29135,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 454
+         }
+      }
+   },
+   "tur-ukr" : {
+      "size_in_bytes" : 384750,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 192375,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2520
+         },
+         "test" : {
+            "num_bytes" : 192375,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2520,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 5040,
+      "config_name" : "tur-ukr",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "ina-lat" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "ina-lat",
+      "dataset_size" : 1824,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 912,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 72447
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 912,
+            "num_bytes" : 72447
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 144894,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "epo-oci" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 54830,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 588,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 54830
+         }
+      },
+      "config_name" : "epo-oci",
+      "dataset_size" : 588,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "eng-kur_Latn" : {
+      "dataset_size" : 290,
+      "config_name" : "eng-kur_Latn",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 18697,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 18697,
+            "num_examples" : 290,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "deu-eng" : {
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 3190940,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 17565,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1595470
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 17565,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1595470
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 35130,
+      "config_name" : "deu-eng",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "jpn_Hira-kor_Hang" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 53616,
+            "name" : "test",
+            "num_examples" : 512,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 53616,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jpn_Hira-kor_Hang",
+      "dataset_size" : 512,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "jbo-rus" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1199,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 109668
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 109668,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jbo-rus",
+      "dataset_size" : 1199,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "swe-cmn_Hant" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "swe-cmn_Hant",
+      "dataset_size" : 567,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 35044,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 567,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 35044,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "ita-yid" : {
+      "config_name" : "ita-yid",
+      "dataset_size" : 412,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 33714,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 206,
+            "name" : "validation",
+            "num_bytes" : 16857
+         },
+         "test" : {
+            "num_bytes" : 16857,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 206,
+            "name" : "test"
+         }
+      }
+   },
+   "ina-nld" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1033,
+      "config_name" : "ina-nld",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 90977,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1033,
+            "num_bytes" : 90977
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "epo-lfn_Latn" : {
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 869,
+            "num_bytes" : 60368
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 869,
+            "num_bytes" : 60368
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 120736,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1738,
+      "config_name" : "epo-lfn_Latn"
+   },
+   "fra-wuu" : {
+      "dataset_size" : 1490,
+      "config_name" : "fra-wuu",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 131616,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 745,
+            "name" : "validation",
+            "num_bytes" : 65808
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 745,
+            "num_bytes" : 65808
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "bul-deu" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 314,
+      "config_name" : "bul-deu",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 314,
+            "num_bytes" : 30294
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 30294,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "fra-tat" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 25778,
+            "name" : "test",
+            "num_examples" : 305,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 25778,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 305,
+      "config_name" : "fra-tat"
+   },
+   "ara-pol" : {
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 16218,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 16218,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 207,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "ara-pol",
+      "dataset_size" : 207,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "nld-ron" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "nld-ron",
+      "dataset_size" : 4538,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 342652,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 171326,
+            "num_examples" : 2269,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 171326,
+            "num_examples" : 2269,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "epo-toki" : {
+      "dataset_size" : 5466,
+      "config_name" : "epo-toki",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 469996,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 234998,
+            "name" : "test",
+            "num_examples" : 2733,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2733,
+            "num_bytes" : 234998
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "bul-zho" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 416,
+      "config_name" : "bul-zho",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 416,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 45135
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 45135,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "rus-uig" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "rus-uig",
+      "dataset_size" : 535,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 535,
+            "name" : "test",
+            "num_bytes" : 66192
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 66192,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "epo-swe" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "epo-swe",
+      "dataset_size" : 3514,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 239432,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1757,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 119716
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1757,
+            "num_bytes" : 119716
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "eng-prg" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 19065,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 19065,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 213,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 213,
+      "config_name" : "eng-prg",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fin-nld" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 317,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 21320
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 21320,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 317,
+      "config_name" : "fin-nld"
+   },
+   "eng-tgl" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 5000,
+      "config_name" : "eng-tgl",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 202000,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500
+         },
+         "validation" : {
+            "num_bytes" : 202000,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2500
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 404000,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "eng-xal" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 281,
+      "config_name" : "eng-xal",
+      "splits" : {
+         "test" : {
+            "num_examples" : 281,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 25020
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 25020,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "vie-vie" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "vie-vie",
+      "dataset_size" : 542,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 542,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 68696
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 68696,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "hun-cmn_Hans" : {
+      "size_in_bytes" : 19862,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 19862,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 247,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "hun-cmn_Hans",
+      "dataset_size" : 247,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "epo-jpn_Hani" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "epo-jpn_Hani",
+      "dataset_size" : 656,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 328,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 31274
+         },
+         "validation" : {
+            "num_bytes" : 31274,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 328,
+            "name" : "validation"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 62548,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "epo-pol" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 5026,
+      "config_name" : "epo-pol",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 381598,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 190799,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2513
+         },
+         "test" : {
+            "num_examples" : 2513,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 190799
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "rus-slv" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1314,
+      "config_name" : "rus-slv",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 113860,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 657,
+            "name" : "validation",
+            "num_bytes" : 56930
+         },
+         "test" : {
+            "num_bytes" : 56930,
+            "num_examples" : 657,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "hrv-ukr" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 32312,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 32312,
+            "name" : "test",
+            "num_examples" : 389,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 389,
+      "config_name" : "hrv-ukr",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fra-lat" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 5830,
+      "config_name" : "fra-lat",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2915,
+            "name" : "test",
+            "num_bytes" : 364086
+         },
+         "validation" : {
+            "num_bytes" : 364086,
+            "name" : "validation",
+            "num_examples" : 2915,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 728172
+   },
+   "fin-jpn_Hira" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 8715,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 841840
+         },
+         "test" : {
+            "num_bytes" : 841840,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 8715
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1683680,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 17430,
+      "config_name" : "fin-jpn_Hira"
+   },
+   "eng-hoc_Latn" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "eng-hoc_Latn",
+      "dataset_size" : 659,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 36698,
+            "name" : "test",
+            "num_examples" : 659,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 36698,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "por-spa" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 1889752,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 944876,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10947,
+            "name" : "validation"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10947,
+            "name" : "test",
+            "num_bytes" : 944876
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 21894,
+      "config_name" : "por-spa",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "deu-est" : {
+      "config_name" : "deu-est",
+      "dataset_size" : 244,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 17015,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 244,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 17015
+         }
+      }
+   },
+   "epo-lit" : {
+      "config_name" : "epo-lit",
+      "dataset_size" : 23022,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1759744,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 879872,
+            "num_examples" : 11511,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 879872,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 11511,
+            "name" : "validation"
+         }
+      }
+   },
+   "ceb-eng" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 23173,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 378,
+            "num_bytes" : 23173
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 378,
+      "config_name" : "ceb-eng",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "kor-zho" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "kor-zho",
+      "dataset_size" : 411,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 37105,
+            "name" : "test",
+            "num_examples" : 411,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 37105,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "lav-rus" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "lav-rus",
+      "dataset_size" : 274,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 22259,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 274,
+            "name" : "test"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 22259
+   },
+   "nob-rus" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1277,
+      "config_name" : "nob-rus",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 1277,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 170914
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 170914,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "cat-deu" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "cat-deu",
+      "dataset_size" : 723,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 59959,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 723,
+            "num_bytes" : 59959
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "jpn_Hani-rus" : {
+      "size_in_bytes" : 201622,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 824,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 100811
+         },
+         "test" : {
+            "num_bytes" : 100811,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 824
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1648,
+      "config_name" : "jpn_Hani-rus",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "ita-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ita-spa",
+      "dataset_size" : 10000,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 763576,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 381788,
+            "num_examples" : 5000,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 381788,
+            "name" : "validation",
+            "num_examples" : 5000,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "epo-ile" : {
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 38878,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 316,
+            "num_bytes" : 19439
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 316,
+            "num_bytes" : 19439
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 632,
+      "config_name" : "epo-ile",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "fin-fin" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1000,
+            "name" : "test",
+            "num_bytes" : 66612
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1000,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 66612
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 133224,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "fin-fin",
+      "dataset_size" : 2000
+   },
+   "ina_Latn-lfn_Latn" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 48484,
+            "num_examples" : 692,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 692,
+            "name" : "test",
+            "num_bytes" : 48484
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 96968,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ina_Latn-lfn_Latn",
+      "dataset_size" : 1384,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "jpn-nor" : {
+      "splits" : {
+         "test" : {
+            "num_bytes" : 104634,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1047
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 104634,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1047,
+      "config_name" : "jpn-nor"
+   },
+   "fra-nor" : {
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "size_in_bytes" : 33674,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 477,
+            "name" : "test",
+            "num_bytes" : 33674
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "fra-nor",
+      "dataset_size" : 477,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "eng-sqi" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 85651,
+            "num_examples" : 1109,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 85651,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-sqi",
+      "dataset_size" : 1109,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "aze_Latn-tur" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1141,
+            "name" : "test",
+            "num_bytes" : 80884
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 80884,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1141,
+      "config_name" : "aze_Latn-tur",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ber-eng" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 21928,
+      "config_name" : "ber-eng",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 896825,
+            "name" : "test",
+            "num_examples" : 10964,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 10964,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 896825
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1793650,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "bel-lat" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 222,
+      "config_name" : "bel-lat",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 222,
+            "name" : "test",
+            "num_bytes" : 19339
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 19339,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "hbs-nor" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 481328,
+            "num_examples" : 5000,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 481328,
+            "num_examples" : 5000,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 962656,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 10000,
+      "config_name" : "hbs-nor",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "spa-tlh" : {
+      "config_name" : "spa-tlh",
+      "dataset_size" : 325,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 20288,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 20288,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 325
+         }
+      },
+      "task_templates" : null
+   },
+   "deu-kor_Hang" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1070,
+      "config_name" : "deu-kor_Hang",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 107534,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1070,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 107534
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "jpn_Hani-cmn_Hans" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 37712,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 18856,
+            "name" : "validation",
+            "num_examples" : 216,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 18856,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 216
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "jpn_Hani-cmn_Hans",
+      "dataset_size" : 432,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "mkd-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "mkd-spa",
+      "dataset_size" : 217,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 14994,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 14994,
+            "num_examples" : 217,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "pol-por" : {
+      "size_in_bytes" : 54377,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 705,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 54377
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 705,
+      "config_name" : "pol-por",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "heb-cmn_Hans" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 285,
+            "num_bytes" : 22756
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 22756,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 285,
+      "config_name" : "heb-cmn_Hans"
+   },
+   "deu-ina" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 2514,
+      "config_name" : "deu-ina",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 137157,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1257,
+            "name" : "validation"
+         },
+         "test" : {
+            "num_bytes" : 137157,
+            "num_examples" : 1257,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 274314,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null
+   },
+   "eng-zsm_Latn" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 49436,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 536
+         },
+         "validation" : {
+            "num_bytes" : 49436,
+            "name" : "validation",
+            "num_examples" : 536,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 98872,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-zsm_Latn",
+      "dataset_size" : 1072,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "hin-zho" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 324,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 44648
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 44648,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 324,
+      "config_name" : "hin-zho"
+   },
+   "nld-cmn_Hant" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1513,
+      "config_name" : "nld-cmn_Hant",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 130492,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1513
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 130492,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "ell-ita" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 36093,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 424,
+            "name" : "test",
+            "num_bytes" : 36093
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "ell-ita",
+      "dataset_size" : 424,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "rus-vie" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 313,
+      "config_name" : "rus-vie",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 34414,
+            "num_examples" : 313,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 34414,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "deu-ita" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 20188,
+      "config_name" : "deu-ita",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 1742628,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 871314,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10094,
+            "name" : "test"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10094,
+            "num_bytes" : 871314
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "nds-nld" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "nds-nld",
+      "dataset_size" : 3314,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 235210,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1657,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 117605
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1657,
+            "name" : "test",
+            "num_bytes" : 117605
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "deu-msa" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 544,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 46985
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 46985,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 544,
+      "config_name" : "deu-msa",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "lat-ukr" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 382,
+      "config_name" : "lat-ukr",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 29975,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 382,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 29975
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "jbo-swe" : {
+      "dataset_size" : 242,
+      "config_name" : "jbo-swe",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 15034,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 15034,
+            "name" : "test",
+            "num_examples" : 242,
+            "dataset_name" : "tatoeba_mt"
+         }
+      }
+   },
+   "dan-nob" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1299,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 101643
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1299,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 101643
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 203286,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "dan-nob",
+      "dataset_size" : 2598
+   },
+   "fin-hun" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1297,
+      "config_name" : "fin-hun",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 80998,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1297,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 80998
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "fin-nno" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 816,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 50880
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 816,
+            "name" : "validation",
+            "num_bytes" : 50880
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 101760,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1632,
+      "config_name" : "fin-nno"
+   },
+   "ces-slv" : {
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 489,
+            "num_bytes" : 28804
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 28804,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ces-slv",
+      "dataset_size" : 489
+   },
+   "ltz-nld" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 17279,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 17279,
+            "num_examples" : 292,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ltz-nld",
+      "dataset_size" : 292,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "por-zho" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "por-zho",
+      "dataset_size" : 1033,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1033,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 75621
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 75621,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "fin-nor" : {
+      "size_in_bytes" : 321226,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 2488,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 160613
+         },
+         "test" : {
+            "num_bytes" : 160613,
+            "name" : "test",
+            "num_examples" : 2488,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fin-nor",
+      "dataset_size" : 4976,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "eng-mkd" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1727104,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 863552,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10010
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10010,
+            "name" : "test",
+            "num_bytes" : 863552
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 20020,
+      "config_name" : "eng-mkd",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-uzb" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 31244,
+            "name" : "test",
+            "num_examples" : 458,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 31244,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-uzb",
+      "dataset_size" : 458,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "cat-spa" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 262524,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1534,
+            "name" : "test",
+            "num_bytes" : 131262
+         },
+         "validation" : {
+            "num_examples" : 1534,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 131262
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "cat-spa",
+      "dataset_size" : 3068,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "epo-rus" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 1248575,
+            "num_examples" : 11674,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 11674,
+            "num_bytes" : 1248575
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2497150,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "epo-rus",
+      "dataset_size" : 23348
+   },
+   "kor-spa" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 83074,
+            "name" : "test",
+            "num_examples" : 940,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 83074,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 940,
+      "config_name" : "kor-spa"
+   },
+   "ita-zho" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 5886,
+      "config_name" : "ita-zho",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2943,
+            "num_bytes" : 217098
+         },
+         "validation" : {
+            "num_bytes" : 217098,
+            "num_examples" : 2943,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 434196,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "eng-mlt" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-mlt",
+      "dataset_size" : 203,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 11648,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 203,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 11648
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "tlh_Latn-cmn_Hans" : {
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 17590,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 246
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 17590,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 246,
+      "config_name" : "tlh_Latn-cmn_Hans",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "jbo-jpn" : {
+      "dataset_size" : 921,
+      "config_name" : "jbo-jpn",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 81231,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 921,
+            "num_bytes" : 81231
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "bel-nld" : {
+      "dataset_size" : 606,
+      "config_name" : "bel-nld",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 62683,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 606,
+            "name" : "test",
+            "num_bytes" : 62683
+         }
+      }
+   },
+   "eng-ron" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 429293,
+            "name" : "validation",
+            "num_examples" : 5508,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 5508,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 429293
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 858586,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-ron",
+      "dataset_size" : 11016,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "bul-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "bul-spa",
+      "dataset_size" : 286,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 24507,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 24507,
+            "num_examples" : 286,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "spa-swe" : {
+      "config_name" : "spa-swe",
+      "dataset_size" : 2702,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "size_in_bytes" : 179774,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 89887,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1351,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 89887,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1351,
+            "name" : "validation"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "eng-kaz_Cyrl" : {
+      "dataset_size" : 396,
+      "config_name" : "eng-kaz_Cyrl",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 36935,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 396,
+            "num_bytes" : 36935
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "spa-toki" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "spa-toki",
+      "dataset_size" : 1280,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 640,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 46015
+         },
+         "validation" : {
+            "num_bytes" : 46015,
+            "name" : "validation",
+            "num_examples" : 640,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 92030,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "eng-kor" : {
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 416066,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 2400,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 208033
+         },
+         "validation" : {
+            "num_bytes" : 208033,
+            "name" : "validation",
+            "num_examples" : 2400,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "eng-kor",
+      "dataset_size" : 4800,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "ara-rus" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "ara-rus",
+      "dataset_size" : 7430,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 366839,
+            "name" : "test",
+            "num_examples" : 3715,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 3715,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 366839
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 733678
+   },
+   "nob-swe" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 563,
+      "config_name" : "nob-swe",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 563,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 38582
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 38582,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "fra-hun" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 353206,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2494,
+            "name" : "validation",
+            "num_bytes" : 176603
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 2494,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 176603
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 4988,
+      "config_name" : "fra-hun",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "ukr-cmn_Hant" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 530,
+      "config_name" : "ukr-cmn_Hant",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 42184,
+            "name" : "test",
+            "num_examples" : 530,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 42184,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "pol-cmn_Hans" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 30094,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 368
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 30094,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 368,
+      "config_name" : "pol-cmn_Hans"
+   },
+   "heb-por" : {
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 719,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 64157
+         },
+         "validation" : {
+            "num_examples" : 719,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 64157
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 128314,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "heb-por",
+      "dataset_size" : 1438
+   },
+   "fra-hrv" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 258,
+      "config_name" : "fra-hrv",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 19608,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 258,
+            "name" : "test",
+            "num_bytes" : 19608
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "srp_Latn-nob" : {
+      "dataset_size" : 9626,
+      "config_name" : "srp_Latn-nob",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 920088,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 4813,
+            "num_bytes" : 460044
+         },
+         "test" : {
+            "num_bytes" : 460044,
+            "num_examples" : 4813,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "bua-rus" : {
+      "dataset_size" : 806,
+      "config_name" : "bua-rus",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 74438,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 806,
+            "num_bytes" : 74438
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "jpn_Hani-spa" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1099,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 102042
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1099,
+            "name" : "validation",
+            "num_bytes" : 102042
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 204084,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jpn_Hani-spa",
+      "dataset_size" : 2198,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "fin-kor_Hang" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 32896,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 32896,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 411,
+            "name" : "test"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fin-kor_Hang",
+      "dataset_size" : 411,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "ara-epo" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ara-epo",
+      "dataset_size" : 1506,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 118142,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 59071,
+            "name" : "validation",
+            "num_examples" : 753,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 59071,
+            "num_examples" : 753,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "ces-ukr" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 3574,
+      "config_name" : "ces-ukr",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 1787,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 127271
+         },
+         "validation" : {
+            "num_bytes" : 127271,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1787,
+            "name" : "validation"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 254542,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "jbo_Latn-cmn_Hant" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "jbo_Latn-cmn_Hant",
+      "dataset_size" : 462,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 15670,
+            "name" : "validation",
+            "num_examples" : 231,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 231,
+            "num_bytes" : 15670
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 31340
+   },
+   "eng-jpn" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 27724,
+      "config_name" : "eng-jpn",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 13862,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1491702
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 13862,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1491702
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 2983404
+   },
+   "aze-spa" : {
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 13994,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 210,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 13994,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 210,
+      "config_name" : "aze-spa",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "deu-fin" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 207955,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2647
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2647,
+            "name" : "test",
+            "num_bytes" : 207955
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 415910,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 5294,
+      "config_name" : "deu-fin",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "epo-cmn_Hant" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 60824,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 831,
+            "name" : "validation"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 831,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 60824
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 121648,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "epo-cmn_Hant",
+      "dataset_size" : 1662,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-tuk" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "eng-tuk",
+      "dataset_size" : 5000,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 213395,
+            "num_examples" : 2500,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 2500,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 213395
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 426790,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "eng-lad" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-lad",
+      "dataset_size" : 1536,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 48380,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 768
+         },
+         "validation" : {
+            "num_examples" : 768,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 48380
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 96760,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "heb-lfn" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 280,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 24022
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 280,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 24022
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 48044,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "heb-lfn",
+      "dataset_size" : 560,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "heb-zho" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 709,
+      "config_name" : "heb-zho",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 709,
+            "num_bytes" : 52399
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 52399,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null
+   },
+   "bel-eng" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 5000,
+      "config_name" : "bel-eng",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500,
+            "num_bytes" : 244579
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 2500,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 244579
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 489158,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "fin-ita" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "fin-ita",
+      "dataset_size" : 2078,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1039,
+            "num_bytes" : 72837
+         },
+         "validation" : {
+            "num_bytes" : 72837,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1039,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 145674
+   },
+   "eng-tlh" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 321225,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 5000,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_examples" : 5000,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 321225
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 642450,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-tlh",
+      "dataset_size" : 10000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "deu-hrv" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "deu-hrv",
+      "dataset_size" : 782,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 58300,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 782,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 58300
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "deu-hun" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "deu-hun",
+      "dataset_size" : 30684,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 15342,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1351044
+         },
+         "validation" : {
+            "num_bytes" : 1351044,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 15342
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 2702088,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "eng-ltz" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-ltz",
+      "dataset_size" : 293,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 18648,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 293,
+            "name" : "test",
+            "num_bytes" : 18648
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "epo-epo" : {
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10000,
+            "num_bytes" : 797530
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 10000,
+            "num_bytes" : 797530
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1595060,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 20000,
+      "config_name" : "epo-epo"
+   },
+   "por-por" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 5000,
+      "config_name" : "por-por",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 409198,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "test",
+            "num_bytes" : 204599
+         },
+         "validation" : {
+            "num_bytes" : 204599,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "validation"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "msa-spa" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "msa-spa",
+      "dataset_size" : 232,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 16593,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 232
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 16593,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "jpn_Hira-nld" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "jpn_Hira-nld",
+      "dataset_size" : 5850,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 269634,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2925
+         },
+         "validation" : {
+            "num_examples" : 2925,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 269634
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 539268,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "eng-gos" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 2308,
+      "config_name" : "eng-gos",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1154,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 56438
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1154,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 56438
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 112876
+   },
+   "fin-fkv" : {
+      "config_name" : "fin-fkv",
+      "dataset_size" : 778,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 64734,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 32367,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 389
+         },
+         "test" : {
+            "num_examples" : 389,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 32367
+         }
+      }
+   },
+   "ina-ita" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 80083,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 912,
+            "name" : "test",
+            "num_bytes" : 80083
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 912,
+      "config_name" : "ina-ita",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-oci" : {
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 54372,
+            "name" : "test",
+            "num_examples" : 841,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 54372,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "eng-oci",
+      "dataset_size" : 841,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ita-por" : {
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 548268,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 3066,
+            "name" : "test",
+            "num_bytes" : 274134
+         },
+         "validation" : {
+            "num_bytes" : 274134,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 3066
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 6132,
+      "config_name" : "ita-por",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "fra-kor_Hang" : {
+      "config_name" : "fra-kor_Hang",
+      "dataset_size" : 316,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 31711,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 316,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 31711
+         }
+      },
+      "task_templates" : null
+   },
+   "eng-lfn_Latn" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 183672,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2450
+         },
+         "test" : {
+            "num_examples" : 2450,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 183672
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 367344,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 4900,
+      "config_name" : "eng-lfn_Latn",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "hun-spa" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "hun-spa",
+      "dataset_size" : 5000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 2500,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 179662
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "validation",
+            "num_bytes" : 179662
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 359324,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null
+   },
+   "pol-zho" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "pol-zho",
+      "dataset_size" : 1004,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 82974,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 82974,
+            "num_examples" : 1004,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "cor-spa" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 11511,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 11511,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 206
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 206,
+      "config_name" : "cor-spa",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-toki" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 802836,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 401418,
+            "num_examples" : 4990,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 401418,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 4990,
+            "name" : "validation"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-toki",
+      "dataset_size" : 9980,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "deu-nor" : {
+      "size_in_bytes" : 747262,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 3651,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 373631
+         },
+         "validation" : {
+            "num_bytes" : 373631,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 3651
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 7302,
+      "config_name" : "deu-nor",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "eng-swe" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-swe",
+      "dataset_size" : 20724,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 697087,
+            "num_examples" : 10362,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 697087,
+            "num_examples" : 10362,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1394174,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "ita-ind" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ita-ind",
+      "dataset_size" : 368,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 27058,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 27058,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 368
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "epo-tgl" : {
+      "splits" : {
+         "test" : {
+            "num_bytes" : 101982,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1109
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 101982,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1109,
+      "config_name" : "epo-tgl"
+   },
+   "nor-nor" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 982,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 67160
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 982,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 67160
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 134320,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1964,
+      "config_name" : "nor-nor"
+   },
+   "ita-cmn_Hans" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ita-cmn_Hans",
+      "dataset_size" : 3654,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 277872,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 138936,
+            "name" : "test",
+            "num_examples" : 1827,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 138936,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1827,
+            "name" : "validation"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "tlh-yid" : {
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 46336,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 23168,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 289,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 23168,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 289
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 578,
+      "config_name" : "tlh-yid",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "fra-ita" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 1436988,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10091,
+            "name" : "validation",
+            "num_bytes" : 718494
+         },
+         "test" : {
+            "num_examples" : 10091,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 718494
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fra-ita",
+      "dataset_size" : 20182,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-jpn_Hani" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 147492,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1500,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_examples" : 1500,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 147492
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 294984,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-jpn_Hani",
+      "dataset_size" : 3000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "jpn-msa" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 485938,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2616,
+            "num_bytes" : 242969
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2616,
+            "name" : "validation",
+            "num_bytes" : 242969
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 5232,
+      "config_name" : "jpn-msa",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fra-msa" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 85373,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 1003,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 85373
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "fra-msa",
+      "dataset_size" : 1003,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "eng-eus" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 167812,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 83906,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1060
+         },
+         "test" : {
+            "num_examples" : 1060,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 83906
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "eng-eus",
+      "dataset_size" : 2120,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "eng-pol" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 1561516,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 780758,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 10099
+         },
+         "validation" : {
+            "num_examples" : 10099,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 780758
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 20198,
+      "config_name" : "eng-pol",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "por-cmn_Hans" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "por-cmn_Hans",
+      "dataset_size" : 638,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 48909,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 638,
+            "num_bytes" : 48909
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      }
+   },
+   "swe-swe" : {
+      "config_name" : "swe-swe",
+      "dataset_size" : 1022,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 72315,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1022,
+            "num_bytes" : 72315
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "hye-rus" : {
+      "config_name" : "hye-rus",
+      "dataset_size" : 227,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 29846,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 227,
+            "name" : "test",
+            "num_bytes" : 29846
+         }
+      }
+   },
+   "hbs-ita" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 534,
+            "name" : "test",
+            "num_bytes" : 37083
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 37083,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 534,
+      "config_name" : "hbs-ita",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-lit" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 5056,
+      "config_name" : "eng-lit",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 180167,
+            "name" : "test",
+            "num_examples" : 2528,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_examples" : 2528,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 180167
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 360334,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "fra-ina" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 2354,
+      "config_name" : "fra-ina",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 210576,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1177,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 105288
+         },
+         "validation" : {
+            "num_bytes" : 105288,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1177
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "afr-deu" : {
+      "config_name" : "afr-deu",
+      "dataset_size" : 1583,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "size_in_bytes" : 99941,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 99941,
+            "num_examples" : 1583,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "ber_Latn-eng" : {
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 896660,
+            "num_examples" : 10962,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_examples" : 10962,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 896660
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1793320,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ber_Latn-eng",
+      "dataset_size" : 21924
+   },
+   "nld-rus" : {
+      "size_in_bytes" : 484888,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2550,
+            "name" : "validation",
+            "num_bytes" : 242444
+         },
+         "test" : {
+            "num_bytes" : 242444,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2550,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "nld-rus",
+      "dataset_size" : 5100,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "eng-ile" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 3422,
+      "config_name" : "eng-ile",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 225104,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1711,
+            "name" : "test",
+            "num_bytes" : 112552
+         },
+         "validation" : {
+            "num_bytes" : 112552,
+            "num_examples" : 1711,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "bul-fra" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 48703,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 446,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 48703
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 446,
+      "config_name" : "bul-fra",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "heb-yid" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 901,
+            "name" : "validation",
+            "num_bytes" : 98653
+         },
+         "test" : {
+            "num_bytes" : 98653,
+            "name" : "test",
+            "num_examples" : 901,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 197306,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "heb-yid",
+      "dataset_size" : 1802,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-ido_Latn" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 1959,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 149043
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 149043,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1959,
+      "config_name" : "eng-ido_Latn"
+   },
+   "eng-tam" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 41917,
+            "name" : "test",
+            "num_examples" : 311,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 41917,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 311,
+      "config_name" : "eng-tam",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "cat-por" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 747,
+      "config_name" : "cat-por",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 65694,
+            "num_examples" : 747,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 65694,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "msa-zho" : {
+      "dataset_size" : 372,
+      "config_name" : "msa-zho",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 33792,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 372,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 33792
+         }
+      },
+      "task_templates" : null
+   },
+   "deu-tat" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 5002,
+      "config_name" : "deu-tat",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 743666,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 2501,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 371833
+         },
+         "validation" : {
+            "num_examples" : 2501,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 371833
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "bel-ita" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "bel-ita",
+      "dataset_size" : 264,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 22431,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 264,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 22431,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "spa-cmn_Hant" : {
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 105555,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1471
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1471,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 105555
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 211110,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "spa-cmn_Hant",
+      "dataset_size" : 2942,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "isl-jpn" : {
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 251,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 21926
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 21926,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "isl-jpn",
+      "dataset_size" : 251,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "run-spa" : {
+      "config_name" : "run-spa",
+      "dataset_size" : 963,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 52277,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 963,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 52277
+         }
+      },
+      "task_templates" : null
+   },
+   "ina-tur" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ina-tur",
+      "dataset_size" : 317,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 20364,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 317,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 20364
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "fin-tur" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1796,
+      "config_name" : "fin-tur",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 140929,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1796,
+            "num_bytes" : 140929
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "eng-mya" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 216,
+      "config_name" : "eng-mya",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 216,
+            "name" : "test",
+            "num_bytes" : 34381
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 34381,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "deu-lat" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "deu-lat",
+      "dataset_size" : 4032,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 285268,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 142634,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2016,
+            "name" : "validation"
+         },
+         "test" : {
+            "num_bytes" : 142634,
+            "num_examples" : 2016,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "heb-spa" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 3698,
+      "config_name" : "heb-spa",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1849,
+            "num_bytes" : 174876
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1849,
+            "num_bytes" : 174876
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 349752,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "cat-fra" : {
+      "dataset_size" : 700,
+      "config_name" : "cat-fra",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 58832,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 700,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 58832
+         }
+      },
+      "task_templates" : null
+   },
+   "eus-spa" : {
+      "size_in_bytes" : 268778,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 134389,
+            "num_examples" : 1850,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 134389,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1850
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eus-spa",
+      "dataset_size" : 3700,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "tur-uig" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1400,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 150099
+         },
+         "test" : {
+            "num_bytes" : 150099,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1400
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 300198,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 2800,
+      "config_name" : "tur-uig",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "jpn_Hani-pol" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 208424,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 104212,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1081
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1081,
+            "name" : "test",
+            "num_bytes" : 104212
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "jpn_Hani-pol",
+      "dataset_size" : 2162,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "deu-nld" : {
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "size_in_bytes" : 1623630,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 10218,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 811815
+         },
+         "test" : {
+            "num_bytes" : 811815,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10218,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 20436,
+      "config_name" : "deu-nld",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "nld-pol" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 2386,
+      "config_name" : "nld-pol",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1193,
+            "name" : "validation",
+            "num_bytes" : 87695
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1193,
+            "num_bytes" : 87695
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 175390
+   },
+   "jpn-mar" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 340,
+      "config_name" : "jpn-mar",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 340,
+            "num_bytes" : 43170
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 43170
+   },
+   "eng-rom" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 706,
+      "config_name" : "eng-rom",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 706,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 48250
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 48250
+   },
+   "ell-nld" : {
+      "size_in_bytes" : 242678,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 922,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 121339
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 922,
+            "name" : "validation",
+            "num_bytes" : 121339
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ell-nld",
+      "dataset_size" : 1844,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "eng-mri" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 366,
+      "config_name" : "eng-mri",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 30911,
+            "name" : "test",
+            "num_examples" : 366,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 30911,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "lfn-rus" : {
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 225,
+            "name" : "test",
+            "num_bytes" : 25529
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 225,
+            "num_bytes" : 25529
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 51058,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "lfn-rus",
+      "dataset_size" : 450
+   },
+   "eng-rus" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-rus",
+      "dataset_size" : 38850,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 3895318,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 19425,
+            "name" : "validation",
+            "num_bytes" : 1947659
+         },
+         "test" : {
+            "num_bytes" : 1947659,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 19425,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "afr-spa" : {
+      "config_name" : "afr-spa",
+      "dataset_size" : 448,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 29333,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 29333,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 448,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null
+   },
+   "rus-ukr" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "rus-ukr",
+      "dataset_size" : 20000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 1107093,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 10000
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 10000,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1107093
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2214186
+   },
+   "cor-deu" : {
+      "size_in_bytes" : 47723,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 821,
+            "name" : "test",
+            "num_bytes" : 47723
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 821,
+      "config_name" : "cor-deu",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "epo-ron" : {
+      "dataset_size" : 7640,
+      "config_name" : "epo-ron",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 559614,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 3820,
+            "num_bytes" : 279807
+         },
+         "validation" : {
+            "num_bytes" : 279807,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 3820
+         }
+      },
+      "task_templates" : null
+   },
+   "pol-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "pol-spa",
+      "dataset_size" : 5088,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 390592,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2544,
+            "num_bytes" : 195296
+         },
+         "validation" : {
+            "num_bytes" : 195296,
+            "num_examples" : 2544,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "spa-tgl" : {
+      "size_in_bytes" : 51801,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 631,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 51801
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 631,
+      "config_name" : "spa-tgl",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "hun-zho" : {
+      "splits" : {
+         "test" : {
+            "num_examples" : 434,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 33218
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 33218,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 434,
+      "config_name" : "hun-zho"
+   },
+   "nld-toki" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 668,
+            "num_bytes" : 43655
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 43655,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 668,
+      "config_name" : "nld-toki"
+   },
+   "kaz_Cyrl-rus" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1398,
+      "config_name" : "kaz_Cyrl-rus",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 337316,
+            "num_examples" : 1398,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 337316
+   },
+   "jpn-tur" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "jpn-tur",
+      "dataset_size" : 811,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 78766,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 811
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 78766
+   },
+   "fra-tur" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 5164,
+      "config_name" : "fra-tur",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 208919,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2582,
+            "name" : "validation"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 2582,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 208919
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 417838,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "deu-jpn_Hira" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 9810,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1070014
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 9810,
+            "num_bytes" : 1070014
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2140028,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 19620,
+      "config_name" : "deu-jpn_Hira"
+   },
+   "bul-cmn_Hans" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 31599,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 31599,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 278,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "bul-cmn_Hans",
+      "dataset_size" : 278,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "tlh-zho" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 448,
+      "config_name" : "tlh-zho",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 31164,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_examples" : 448,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 31164
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      }
+   },
+   "ina_Latn-lad_Latn" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "ina_Latn-lad_Latn",
+      "dataset_size" : 464,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 16210,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 232
+         },
+         "test" : {
+            "num_bytes" : 16210,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 232,
+            "name" : "test"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 32420,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "eng-uzb_Latn" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-uzb_Latn",
+      "dataset_size" : 301,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 301,
+            "num_bytes" : 17949
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 17949,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "kor_Hang-spa" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 921,
+      "config_name" : "kor_Hang-spa",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 921,
+            "name" : "test",
+            "num_bytes" : 81877
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 81877,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "ita-toki" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 200,
+      "config_name" : "ita-toki",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 17073,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 17073,
+            "name" : "test",
+            "num_examples" : 200,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "por-swe" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "por-swe",
+      "dataset_size" : 320,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 21549,
+            "num_examples" : 320,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 21549,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "swe-cmn_Hans" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 33354,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 478,
+            "name" : "test",
+            "num_bytes" : 33354
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "swe-cmn_Hans",
+      "dataset_size" : 478,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "bul-rus" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "bul-rus",
+      "dataset_size" : 2494,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 285170,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1247,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 142585
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1247,
+            "name" : "validation",
+            "num_bytes" : 142585
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "epo-yid" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1986,
+      "config_name" : "epo-yid",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 993,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 94278
+         },
+         "test" : {
+            "num_bytes" : 94278,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 993
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 188556,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "ota-tur" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "ota-tur",
+      "dataset_size" : 614,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 29366,
+            "name" : "validation",
+            "num_examples" : 307,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 29366,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 307
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 58732,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "ara-spa" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 237702,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1511,
+            "name" : "validation",
+            "num_bytes" : 118851
+         },
+         "test" : {
+            "num_bytes" : 118851,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1511,
+            "name" : "test"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 3022,
+      "config_name" : "ara-spa",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "por-toki" : {
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 159364,
+            "num_examples" : 1719,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 159364,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1719
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 318728,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 3438,
+      "config_name" : "por-toki",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-ind" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 659220,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 4289,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 329610
+         },
+         "validation" : {
+            "num_bytes" : 329610,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 4289,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "eng-ind",
+      "dataset_size" : 8578,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "deu-heb" : {
+      "config_name" : "deu-heb",
+      "dataset_size" : 6180,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 610250,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 305125,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 3090
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 3090,
+            "num_bytes" : 305125
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "eng-kha" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 84084,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1314
+         },
+         "test" : {
+            "num_bytes" : 84084,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1314
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 168168,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-kha",
+      "dataset_size" : 2628
+   },
+   "ita-swe" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ita-swe",
+      "dataset_size" : 715,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 715,
+            "num_bytes" : 51351
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 51351,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "eng-pes" : {
+      "config_name" : "eng-pes",
+      "dataset_size" : 3757,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 409230,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 3757,
+            "num_bytes" : 409230
+         }
+      }
+   },
+   "ita-pol" : {
+      "dataset_size" : 2588,
+      "config_name" : "ita-pol",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 188026,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 94013,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1294,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 94013,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1294
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "ita-jpn_Hani" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "ita-jpn_Hani",
+      "dataset_size" : 576,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 288,
+            "name" : "validation",
+            "num_bytes" : 27233
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 288,
+            "name" : "test",
+            "num_bytes" : 27233
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 54466,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "eng-hye" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 2242,
+      "config_name" : "eng-hye",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 146738,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1121,
+            "name" : "test",
+            "num_bytes" : 73369
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1121,
+            "name" : "validation",
+            "num_bytes" : 73369
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      }
+   },
+   "arq-eng" : {
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 405,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 36058
+         },
+         "validation" : {
+            "num_bytes" : 36058,
+            "name" : "validation",
+            "num_examples" : 405,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 72116,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 810,
+      "config_name" : "arq-eng",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "spa-zho" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 195859,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2705,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 195859,
+            "name" : "validation",
+            "num_examples" : 2705,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 391718,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 5410,
+      "config_name" : "spa-zho",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-cmn_Hans" : {
+      "size_in_bytes" : 771268,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 385634,
+            "num_examples" : 4450,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 4450,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 385634
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 8900,
+      "config_name" : "eng-cmn_Hans",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "epo-spa" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 21470,
+      "config_name" : "epo-spa",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 894253,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 10735
+         },
+         "validation" : {
+            "num_bytes" : 894253,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10735,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1788506,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "kor-rus" : {
+      "size_in_bytes" : 23083,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 221,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 23083
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 221,
+      "config_name" : "kor-rus",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "fra-vie" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 136035,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1299
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 136035,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "fra-vie",
+      "dataset_size" : 1299
+   },
+   "jpn-vie" : {
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 245620,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 122810,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1097
+         },
+         "validation" : {
+            "num_bytes" : 122810,
+            "num_examples" : 1097,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 2194,
+      "config_name" : "jpn-vie",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "dtp-jpn" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 251,
+      "config_name" : "dtp-jpn",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 21300,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 251,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 21300,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "ita-lit" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 16502,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 16502,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 224,
+            "name" : "test"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 224,
+      "config_name" : "ita-lit",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "ces-lat" : {
+      "splits" : {
+         "test" : {
+            "num_bytes" : 22590,
+            "name" : "test",
+            "num_examples" : 323,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 22590,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ces-lat",
+      "dataset_size" : 323
+   },
+   "lad_Latn-yid" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 37993,
+            "name" : "validation",
+            "num_examples" : 438,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 37993,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 438,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 75986,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "lad_Latn-yid",
+      "dataset_size" : 876
+   },
+   "dan-jpn_Hira" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "splits" : {
+         "test" : {
+            "num_examples" : 716,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 59987
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 59987,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 716,
+      "config_name" : "dan-jpn_Hira",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "dan-dan" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 212,
+      "config_name" : "dan-dan",
+      "splits" : {
+         "test" : {
+            "num_examples" : 212,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 18075
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 18075,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "ita-jpn" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 547232,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 273616,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2660
+         },
+         "test" : {
+            "num_examples" : 2660,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 273616
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "ita-jpn",
+      "dataset_size" : 5320,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "epo-ido" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 2366,
+      "config_name" : "epo-ido",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 174618,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 87309,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1183,
+            "name" : "validation"
+         },
+         "test" : {
+            "num_bytes" : 87309,
+            "name" : "test",
+            "num_examples" : 1183,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "eng-gla" : {
+      "size_in_bytes" : 68841,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 955,
+            "num_bytes" : 68841
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-gla",
+      "dataset_size" : 955,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "jpn_Hira-jpn_Hira" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jpn_Hira-jpn_Hira",
+      "dataset_size" : 441,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 60247,
+            "num_examples" : 441,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 60247,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "eng-kab" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 24284,
+      "config_name" : "eng-kab",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1625614,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 812807,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 12142,
+            "name" : "validation"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 12142,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 812807
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "jpn_Hira-ind" : {
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 2256,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 209281
+         },
+         "test" : {
+            "num_bytes" : 209281,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2256,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 418562,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jpn_Hira-ind",
+      "dataset_size" : 4512
+   },
+   "slv-ukr" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1830,
+      "config_name" : "slv-ukr",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 60177,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 915
+         },
+         "test" : {
+            "num_examples" : 915,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 60177
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 120354
+   },
+   "eng-fry" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 15737,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 220,
+            "num_bytes" : 15737
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 220,
+      "config_name" : "eng-fry",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "epo-ile_Latn" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 19423,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 315,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 19423
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 315,
+      "config_name" : "epo-ile_Latn",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "fra-pcd" : {
+      "dataset_size" : 266,
+      "config_name" : "fra-pcd",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "size_in_bytes" : 18060,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 18060,
+            "num_examples" : 266,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "eng-fra" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-fra",
+      "dataset_size" : 25362,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 1106067,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 12681
+         },
+         "test" : {
+            "num_bytes" : 1106067,
+            "name" : "test",
+            "num_examples" : 12681,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2212134,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "cor-epo" : {
+      "config_name" : "cor-epo",
+      "dataset_size" : 663,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 36954,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 663,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 36954
+         }
+      }
+   },
+   "ido-ina" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ido-ina",
+      "dataset_size" : 632,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 20767,
+            "name" : "test",
+            "num_examples" : 316,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_examples" : 316,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 20767
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 41534,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "lit-pol" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 126233,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1787,
+            "name" : "test"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 126233,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "lit-pol",
+      "dataset_size" : 1787
+   },
+   "fra-hbs" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 474,
+            "name" : "test",
+            "num_bytes" : 37837
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 37837,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "fra-hbs",
+      "dataset_size" : 474,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "dan-nld" : {
+      "dataset_size" : 3286,
+      "config_name" : "dan-nld",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "size_in_bytes" : 218486,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 109243,
+            "num_examples" : 1643,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1643,
+            "num_bytes" : 109243
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "ido-ita" : {
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 228410,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 114205,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1460
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1460,
+            "num_bytes" : 114205
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 2920,
+      "config_name" : "ido-ita",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "eng-por" : {
+      "dataset_size" : 26444,
+      "config_name" : "eng-por",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2170336,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 1085168,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 13222
+         },
+         "test" : {
+            "num_bytes" : 1085168,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 13222
+         }
+      },
+      "task_templates" : null
+   },
+   "lfn-por" : {
+      "dataset_size" : 3890,
+      "config_name" : "lfn-por",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 561430,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 280715,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1945,
+            "name" : "validation"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1945,
+            "num_bytes" : 280715
+         }
+      },
+      "task_templates" : null
+   },
+   "gos-nld" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 3704,
+      "config_name" : "gos-nld",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 200818,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 1852,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 100409
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1852,
+            "name" : "test",
+            "num_bytes" : 100409
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "jbo-zho" : {
+      "size_in_bytes" : 72678,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 518,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 36339
+         },
+         "validation" : {
+            "num_examples" : 518,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 36339
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1036,
+      "config_name" : "jbo-zho",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "deu-ukr" : {
+      "config_name" : "deu-ukr",
+      "dataset_size" : 20638,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1728644,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 10319,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 864322
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10319,
+            "name" : "test",
+            "num_bytes" : 864322
+         }
+      },
+      "task_templates" : null
+   },
+   "hun-kor" : {
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 22711,
+            "num_examples" : 271,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 22711,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 271,
+      "config_name" : "hun-kor",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "srp_Latn-rus" : {
+      "config_name" : "srp_Latn-rus",
+      "dataset_size" : 2966,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "size_in_bytes" : 254596,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1483,
+            "name" : "test",
+            "num_bytes" : 127298
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1483,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 127298
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "tlh_Latn-cmn_Hant" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "size_in_bytes" : 13574,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 13574,
+            "num_examples" : 202,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 202,
+      "config_name" : "tlh_Latn-cmn_Hant",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "por-rus" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "por-rus",
+      "dataset_size" : 20000,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2024516,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 10000,
+            "num_bytes" : 1012258
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10000,
+            "num_bytes" : 1012258
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "nor-ukr" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 59241,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 59241,
+            "name" : "test",
+            "num_examples" : 670,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 670,
+      "config_name" : "nor-ukr",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "spa-spa" : {
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "test",
+            "num_bytes" : 231019
+         },
+         "validation" : {
+            "num_bytes" : 231019,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2500
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 462038,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "spa-spa",
+      "dataset_size" : 5000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-war" : {
+      "dataset_size" : 1512,
+      "config_name" : "eng-war",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 117917,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1512,
+            "name" : "test",
+            "num_bytes" : 117917
+         }
+      }
+   },
+   "heb-cmn_Hant" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "heb-cmn_Hant",
+      "dataset_size" : 329,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 23289,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 329,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 23289,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "nob-spa" : {
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 885,
+            "name" : "test",
+            "num_bytes" : 74321
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 74321,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 885,
+      "config_name" : "nob-spa"
+   },
+   "eng-mon" : {
+      "size_in_bytes" : 92558,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 46279,
+            "num_examples" : 415,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 415,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 46279
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 830,
+      "config_name" : "eng-mon",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "epo-zho" : {
+      "size_in_bytes" : 345134,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2241,
+            "num_bytes" : 172567
+         },
+         "test" : {
+            "num_examples" : 2241,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 172567
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 4482,
+      "config_name" : "epo-zho",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "fin-nob" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 219466,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1672,
+            "name" : "test",
+            "num_bytes" : 109733
+         },
+         "validation" : {
+            "num_bytes" : 109733,
+            "name" : "validation",
+            "num_examples" : 1672,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fin-nob",
+      "dataset_size" : 3344,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "rus-uig_Arab" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 65448,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 529
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 65448,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "rus-uig_Arab",
+      "dataset_size" : 529
+   },
+   "epo-lfn" : {
+      "dataset_size" : 1994,
+      "config_name" : "epo-lfn",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 147258,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 997,
+            "num_bytes" : 73629
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 997,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 73629
+         }
+      },
+      "task_templates" : null
+   },
+   "epo-isl" : {
+      "config_name" : "epo-isl",
+      "dataset_size" : 233,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 17137,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 17137,
+            "name" : "test",
+            "num_examples" : 233,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null
+   },
+   "ita-rus" : {
+      "dataset_size" : 20090,
+      "config_name" : "ita-rus",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "size_in_bytes" : 1988900,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 994450,
+            "name" : "validation",
+            "num_examples" : 10045,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10045,
+            "name" : "test",
+            "num_bytes" : 994450
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "avk-spa" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 275,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 18491
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 18491,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 275,
+      "config_name" : "avk-spa"
+   },
+   "tur-tur" : {
+      "dataset_size" : 5000,
+      "config_name" : "tur-tur",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 364230,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2500,
+            "num_bytes" : 182115
+         },
+         "test" : {
+            "num_bytes" : 182115,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500
+         }
+      },
+      "task_templates" : null
+   },
+   "eng-zza" : {
+      "size_in_bytes" : 29734,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 29734,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 529
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 529,
+      "config_name" : "eng-zza",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "rus-yue_Hans" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "rus-yue_Hans",
+      "dataset_size" : 448,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 23533,
+            "num_examples" : 224,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 23533,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 224,
+            "name" : "validation"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 47066
+   },
+   "dan-fin" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "dan-fin",
+      "dataset_size" : 5330,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 357738,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 178869,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2665
+         },
+         "validation" : {
+            "num_bytes" : 178869,
+            "num_examples" : 2665,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "ces-hun" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 121638,
+            "name" : "test",
+            "num_examples" : 1911,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1911,
+            "name" : "validation",
+            "num_bytes" : 121638
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 243276,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 3822,
+      "config_name" : "ces-hun"
+   },
+   "hsb-slv" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 743,
+      "config_name" : "hsb-slv",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 35522,
+            "name" : "test",
+            "num_examples" : 743,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 35522,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "bre-eng" : {
+      "splits" : {
+         "test" : {
+            "num_bytes" : 21410,
+            "name" : "test",
+            "num_examples" : 383,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 21410,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "bre-eng",
+      "dataset_size" : 383
+   },
+   "spa-yid" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 74990,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 37495,
+            "name" : "test",
+            "num_examples" : 407,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 407,
+            "num_bytes" : 37495
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "spa-yid",
+      "dataset_size" : 814,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "cha-eng" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 227,
+      "config_name" : "cha-eng",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 11634,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 227,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 11634
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "ara-deu" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 2418,
+      "config_name" : "ara-deu",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 205344,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 1209,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 102672
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1209,
+            "name" : "validation",
+            "num_bytes" : 102672
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      }
+   },
+   "eng-hoc" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 36715,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 36715,
+            "num_examples" : 660,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "eng-hoc",
+      "dataset_size" : 660,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "chv-rus" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 382,
+      "config_name" : "chv-rus",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 40925,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 40925,
+            "num_examples" : 382,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "rus-tat" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "rus-tat",
+      "dataset_size" : 4276,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 448786,
+            "num_examples" : 2138,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_examples" : 2138,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 448786
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 897572,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "pol-cmn_Hant" : {
+      "dataset_size" : 596,
+      "config_name" : "pol-cmn_Hant",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 50414,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 50414,
+            "name" : "test",
+            "num_examples" : 596,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null
+   },
+   "ukr-cmn_Hans" : {
+      "dataset_size" : 853,
+      "config_name" : "ukr-cmn_Hans",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "size_in_bytes" : 79448,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 79448,
+            "num_examples" : 853,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "epo-lad_Latn" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "epo-lad_Latn",
+      "dataset_size" : 528,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 264,
+            "num_bytes" : 17121
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 264,
+            "name" : "test",
+            "num_bytes" : 17121
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 34242
+   },
+   "afr-epo" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1116,
+            "name" : "test",
+            "num_bytes" : 81498
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 81498,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "afr-epo",
+      "dataset_size" : 1116,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "fra-gcf" : {
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 69687,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 69687,
+            "num_examples" : 1164,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 1164,
+      "config_name" : "fra-gcf",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "fra-nob" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 323,
+      "config_name" : "fra-nob",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 323,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 24303
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 24303,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "fra-uig" : {
+      "dataset_size" : 694,
+      "config_name" : "fra-uig",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "size_in_bytes" : 66867,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 66867,
+            "name" : "test",
+            "num_examples" : 694,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "eng-jbo" : {
+      "size_in_bytes" : 728086,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 364043,
+            "num_examples" : 4996,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_examples" : 4996,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 364043
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-jbo",
+      "dataset_size" : 9992,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "bul-jpn" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "bul-jpn",
+      "dataset_size" : 321,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 39835,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 321,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 39835,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "lit-rus" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "lit-rus",
+      "dataset_size" : 7196,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 3598,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 328689
+         },
+         "test" : {
+            "num_examples" : 3598,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 328689
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 657378,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "deu-jpn_Kana" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 33355,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 321
+         },
+         "validation" : {
+            "num_bytes" : 33355,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 321,
+            "name" : "validation"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 66710,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "deu-jpn_Kana",
+      "dataset_size" : 642
+   },
+   "jbo-spa" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "jbo-spa",
+      "dataset_size" : 1507,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 107509,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1507,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 107509,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "deu-nds" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 19998,
+      "config_name" : "deu-nds",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 9999,
+            "num_bytes" : 796333
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 9999,
+            "name" : "test",
+            "num_bytes" : 796333
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1592666,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "nld-por" : {
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 185289,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2500
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500,
+            "num_bytes" : 185289
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 370578,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 5000,
+      "config_name" : "nld-por"
+   },
+   "lat_Latn-por" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 5000,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 625922
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 625922,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "lat_Latn-por",
+      "dataset_size" : 5000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-kzj" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1169,
+            "name" : "test",
+            "num_bytes" : 107218
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 107218,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "eng-kzj",
+      "dataset_size" : 1169,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "jpn_Hira-por" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 3080,
+      "config_name" : "jpn_Hira-por",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 314524,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1540,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 157262
+         },
+         "test" : {
+            "num_bytes" : 157262,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1540,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "dan-nor" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 102342,
+            "num_examples" : 1311,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1311,
+            "name" : "validation",
+            "num_bytes" : 102342
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 204684,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "dan-nor",
+      "dataset_size" : 2622,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "srp_Cyrl-ukr" : {
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 19037,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 205,
+            "name" : "test",
+            "num_bytes" : 19037
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 205,
+      "config_name" : "srp_Cyrl-ukr",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "fra-slv" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 38259,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 448,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 38259
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "fra-slv",
+      "dataset_size" : 448,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "eng-kur" : {
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 61790,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 442,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 30895
+         },
+         "test" : {
+            "num_bytes" : 30895,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 442,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "eng-kur",
+      "dataset_size" : 884,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "kzj_Latn-zsm_Latn" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 43012,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 347,
+            "name" : "test",
+            "num_bytes" : 43012
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "kzj_Latn-zsm_Latn",
+      "dataset_size" : 347,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-tzl" : {
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 10369,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 201,
+            "name" : "validation"
+         },
+         "test" : {
+            "num_bytes" : 10369,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 201,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 20738,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 402,
+      "config_name" : "eng-tzl"
+   },
+   "eng-fas" : {
+      "size_in_bytes" : 820830,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 410415,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 3762
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 3762,
+            "name" : "validation",
+            "num_bytes" : 410415
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 7524,
+      "config_name" : "eng-fas",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "eng-run" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 90358,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1703,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 90358,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1703,
+      "config_name" : "eng-run",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-ilo" : {
+      "size_in_bytes" : 156808,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1093,
+            "name" : "validation",
+            "num_bytes" : 78404
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1093,
+            "num_bytes" : 78404
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-ilo",
+      "dataset_size" : 2186,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "tur-uzb" : {
+      "size_in_bytes" : 14378,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 208,
+            "num_bytes" : 14378
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "tur-uzb",
+      "dataset_size" : 208,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "dan-rus" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 3426,
+      "config_name" : "dan-rus",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 313060,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1713,
+            "num_bytes" : 156530
+         },
+         "validation" : {
+            "num_bytes" : 156530,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1713
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "ukr-ukr" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 824,
+      "config_name" : "ukr-ukr",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 73415,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 824
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 73415,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "tat-tur" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 521,
+      "config_name" : "tat-tur",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 46968,
+            "name" : "test",
+            "num_examples" : 521,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 46968,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "ita-pms" : {
+      "config_name" : "ita-pms",
+      "dataset_size" : 232,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "size_in_bytes" : 18864,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 18864,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 232,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "epo-srp_Latn" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 20921,
+            "name" : "validation",
+            "num_examples" : 270,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 20921,
+            "name" : "test",
+            "num_examples" : 270,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 41842,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 540,
+      "config_name" : "epo-srp_Latn"
+   },
+   "deu-por" : {
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 10000,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 894266
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10000,
+            "name" : "test",
+            "num_bytes" : 894266
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1788532,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "deu-por",
+      "dataset_size" : 20000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "nor-por" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "nor-por",
+      "dataset_size" : 481,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 43546,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 481
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 43546,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "rus-tlh" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "rus-tlh",
+      "dataset_size" : 254,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 19143,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 254,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 19143
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "ell-por" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ell-por",
+      "dataset_size" : 1770,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 144272,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 885,
+            "name" : "validation",
+            "num_bytes" : 72136
+         },
+         "test" : {
+            "num_bytes" : 72136,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 885,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "cor-eng" : {
+      "config_name" : "cor-eng",
+      "dataset_size" : 6396,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 327494,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 163747,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 3198
+         },
+         "test" : {
+            "num_bytes" : 163747,
+            "name" : "test",
+            "num_examples" : 3198,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null
+   },
+   "jpn-zho" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 4994,
+      "config_name" : "jpn-zho",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2497,
+            "name" : "test",
+            "num_bytes" : 244984
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2497,
+            "name" : "validation",
+            "num_bytes" : 244984
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 489968,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "fra-zho" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 19988,
+      "config_name" : "fra-zho",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 896410,
+            "name" : "test",
+            "num_examples" : 9994,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 896410,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 9994,
+            "name" : "validation"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1792820,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "nds-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "nds-spa",
+      "dataset_size" : 923,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 59877,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 59877,
+            "num_examples" : 923,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "fra-lfn" : {
+      "config_name" : "fra-lfn",
+      "dataset_size" : 1046,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 111380,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 55690,
+            "name" : "test",
+            "num_examples" : 523,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 55690,
+            "num_examples" : 523,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      }
+   },
+   "spa-vie" : {
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 594,
+            "num_bytes" : 46565
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 46565,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "spa-vie",
+      "dataset_size" : 594,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ell-fra" : {
+      "config_name" : "ell-fra",
+      "dataset_size" : 1506,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 135841,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 135841,
+            "name" : "test",
+            "num_examples" : 1506,
+            "dataset_name" : "tatoeba_mt"
+         }
+      }
+   },
+   "deu-kab" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 373,
+      "config_name" : "deu-kab",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 25648,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 373,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 25648
+   },
+   "ell-ell" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ell-ell",
+      "dataset_size" : 533,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 57805,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 57805,
+            "num_examples" : 533,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "deu-fra" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 24836,
+      "config_name" : "deu-fra",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 2266172,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 1133086,
+            "name" : "validation",
+            "num_examples" : 12418,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_examples" : 12418,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1133086
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "hbs-zho" : {
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 17387,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 236,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 17387,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 236,
+      "config_name" : "hbs-zho",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "deu-ell" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 477796,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500,
+            "num_bytes" : 238898
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "validation",
+            "num_bytes" : 238898
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 5000,
+      "config_name" : "deu-ell",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "hun-tur" : {
+      "dataset_size" : 1002,
+      "config_name" : "hun-tur",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 63469,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 63469,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1002,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "ces-rus" : {
+      "config_name" : "ces-rus",
+      "dataset_size" : 5868,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 508254,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 254127,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2934
+         },
+         "validation" : {
+            "num_bytes" : 254127,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2934,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null
+   },
+   "cat-nld" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 41913,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 578,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 41913
+         }
+      },
+      "dataset_size" : 578,
+      "config_name" : "cat-nld",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "heb-ina" : {
+      "dataset_size" : 434,
+      "config_name" : "heb-ina",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 38664,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 19332,
+            "name" : "test",
+            "num_examples" : 217,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 217,
+            "num_bytes" : 19332
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "deu-cmn_Hans" : {
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2609,
+            "num_bytes" : 227663
+         },
+         "test" : {
+            "num_examples" : 2609,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 227663
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 455326,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "deu-cmn_Hans",
+      "dataset_size" : 5218,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ron-rus" : {
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 66481,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 782,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 66481,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 782,
+      "config_name" : "ron-rus",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "nld-ukr" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 775813,
+            "name" : "validation",
+            "num_examples" : 10000,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 775813,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 10000
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1551626,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "nld-ukr",
+      "dataset_size" : 20000
+   },
+   "bul-jpn_Hira" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 33801,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 264,
+            "name" : "test"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 33801,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 264,
+      "config_name" : "bul-jpn_Hira"
+   },
+   "fra-tlh_Latn" : {
+      "config_name" : "fra-tlh_Latn",
+      "dataset_size" : 619,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 41512,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 41512,
+            "num_examples" : 619,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null
+   },
+   "dsb-hsb" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 331,
+      "config_name" : "dsb-hsb",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 15387,
+            "name" : "test",
+            "num_examples" : 331,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 15387,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "epo-slv" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 302,
+            "name" : "test",
+            "num_bytes" : 23153
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 23153,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 302,
+      "config_name" : "epo-slv"
+   },
+   "heb-ita" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 285968,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 142984,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1706
+         },
+         "validation" : {
+            "num_bytes" : 142984,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1706,
+            "name" : "validation"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 3412,
+      "config_name" : "heb-ita",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-orv" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 26658,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 26658,
+            "name" : "test",
+            "num_examples" : 322,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 322,
+      "config_name" : "eng-orv",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "yid-yid" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "yid-yid",
+      "dataset_size" : 292,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 35801,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 292,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 35801
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "tlh_Latn-yid" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 23118,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 288
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 23118,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "tlh_Latn-yid",
+      "dataset_size" : 288,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "aze-eng" : {
+      "dataset_size" : 5318,
+      "config_name" : "aze-eng",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "size_in_bytes" : 370942,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 185471,
+            "name" : "test",
+            "num_examples" : 2659,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 185471,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2659,
+            "name" : "validation"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "deu-hsb" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 666,
+      "config_name" : "deu-hsb",
+      "splits" : {
+         "test" : {
+            "num_examples" : 666,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 50779
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 50779,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "epo-nob" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "epo-nob",
+      "dataset_size" : 6184,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 3092,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 308406
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 3092,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 308406
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 616812,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "fin-zho" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 382,
+      "config_name" : "fin-zho",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 24985,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 382
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 24985,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "rus-swe" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 2564,
+      "config_name" : "rus-swe",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 233246,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1282,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 116623
+         },
+         "test" : {
+            "num_examples" : 1282,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 116623
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "ina-lfn" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 1538,
+      "config_name" : "ina-lfn",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 769,
+            "num_bytes" : 55918
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 769,
+            "name" : "test",
+            "num_bytes" : 55918
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 111836
+   },
+   "jpn_Hira-nob" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 805,
+      "config_name" : "jpn_Hira-nob",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 805,
+            "name" : "test",
+            "num_bytes" : 83793
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 83793,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "aze-tur" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "aze-tur",
+      "dataset_size" : 1150,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 1150,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 81567
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 81567,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "rus-toki" : {
+      "dataset_size" : 1800,
+      "config_name" : "rus-toki",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 173216,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 86608,
+            "name" : "validation",
+            "num_examples" : 900,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 86608,
+            "name" : "test",
+            "num_examples" : 900,
+            "dataset_name" : "tatoeba_mt"
+         }
+      }
+   },
+   "bel-spa" : {
+      "dataset_size" : 205,
+      "config_name" : "bel-spa",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 18897,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 205,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 18897
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "ita-nor" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 102588,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 102588,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 939
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 939,
+      "config_name" : "ita-nor",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fry-nld" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 18557,
+            "num_examples" : 260,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 18557,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "fry-nld",
+      "dataset_size" : 260
+   },
+   "eng-heb" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 1806294,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 10519,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 903147
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10519,
+            "name" : "validation",
+            "num_bytes" : 903147
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 21038,
+      "config_name" : "eng-heb",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "deu-ind" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 497,
+      "config_name" : "deu-ind",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 42298,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 42298,
+            "num_examples" : 497,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "eng-srp_Cyrl" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 273082,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 136541,
+            "name" : "test",
+            "num_examples" : 1580,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 136541,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1580,
+            "name" : "validation"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-srp_Cyrl",
+      "dataset_size" : 3160,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "dan-swe" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 3098,
+      "config_name" : "dan-swe",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 107366,
+            "num_examples" : 1549,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 107366,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1549
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 214732,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "ces-pol" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 40371,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 568
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 40371,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 568,
+      "config_name" : "ces-pol"
+   },
+   "deu-kur" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 237,
+      "config_name" : "deu-kur",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 16071,
+            "num_examples" : 237,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 16071,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "jpn-spa" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 10742,
+            "num_bytes" : 1065342
+         },
+         "validation" : {
+            "num_bytes" : 1065342,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10742
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 2130684,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jpn-spa",
+      "dataset_size" : 21484,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "fra-spa" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 883143,
+            "name" : "test",
+            "num_examples" : 10294,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10294,
+            "num_bytes" : 883143
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 1766286,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 20588,
+      "config_name" : "fra-spa",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "hun-kor_Hang" : {
+      "config_name" : "hun-kor_Hang",
+      "dataset_size" : 268,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 22473,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 268,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 22473
+         }
+      }
+   },
+   "eng-urd" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1663,
+      "config_name" : "eng-urd",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 160779,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 160779,
+            "name" : "test",
+            "num_examples" : 1663,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "epo-vie" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1790,
+      "config_name" : "epo-vie",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 173445,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 173445,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1790
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "ber-fra" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 10144,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 758124
+         },
+         "test" : {
+            "num_bytes" : 758124,
+            "name" : "test",
+            "num_examples" : 10144,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1516248,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ber-fra",
+      "dataset_size" : 20288,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "cym-eng" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 818,
+            "num_bytes" : 57737
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 57737,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "cym-eng",
+      "dataset_size" : 818,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "deu-fas" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "deu-fas",
+      "dataset_size" : 6370,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 362389,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 3185
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 3185,
+            "num_bytes" : 362389
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 724778,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "slv-cmn_Hans" : {
+      "dataset_size" : 718,
+      "config_name" : "slv-cmn_Hans",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 54243,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 54243,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 718
+         }
+      }
+   },
+   "deu-run" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 100540,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1752
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1752,
+            "num_bytes" : 100540
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 201080,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "deu-run",
+      "dataset_size" : 3504,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "khm-spa" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 147017,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1448,
+            "num_bytes" : 147017
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1448,
+      "config_name" : "khm-spa",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "hbs-spa" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 42872,
+            "name" : "test",
+            "num_examples" : 607,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 42872,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 607,
+      "config_name" : "hbs-spa",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "pol-tur" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "pol-tur",
+      "dataset_size" : 892,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 66267,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 892,
+            "name" : "test"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 66267,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "ita-lat" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 3432,
+      "config_name" : "ita-lat",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 1716,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 132924
+         },
+         "test" : {
+            "num_bytes" : 132924,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1716,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 265848
+   },
+   "cor-ita" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 287,
+      "config_name" : "cor-ita",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 287,
+            "name" : "test",
+            "num_bytes" : 14528
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 14528
+   },
+   "eng-lav" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 114179,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1631
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 114179,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "eng-lav",
+      "dataset_size" : 1631
+   },
+   "dtp-jpn_Hira" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 20048,
+            "num_examples" : 236,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 20048,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "dtp-jpn_Hira",
+      "dataset_size" : 236,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "hun-ita" : {
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 329245,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 5000
+         },
+         "test" : {
+            "num_bytes" : 329245,
+            "name" : "test",
+            "num_examples" : 5000,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 658490,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "hun-ita",
+      "dataset_size" : 10000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "jpn_Hani-por" : {
+      "splits" : {
+         "test" : {
+            "num_examples" : 336,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 30661
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 336,
+            "name" : "validation",
+            "num_bytes" : 30661
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 61322,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "jpn_Hani-por",
+      "dataset_size" : 672
+   },
+   "afr-eng" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 96460,
+            "name" : "validation",
+            "num_examples" : 1374,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1374,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 96460
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 192920,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 2748,
+      "config_name" : "afr-eng"
+   },
+   "fra-yid" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 384,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 35558
+         },
+         "validation" : {
+            "num_bytes" : 35558,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 384,
+            "name" : "validation"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 71116,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "fra-yid",
+      "dataset_size" : 768,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "deu-jbo" : {
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1449,
+            "name" : "test",
+            "num_bytes" : 114176
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 114176,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "deu-jbo",
+      "dataset_size" : 1449
+   },
+   "jpn_Hira-lit" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 205,
+      "config_name" : "jpn_Hira-lit",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 18266,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 205,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 18266,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      }
+   },
+   "jpn_Hira-cmn_Hant" : {
+      "config_name" : "jpn_Hira-cmn_Hant",
+      "dataset_size" : 1364,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 131750,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 682,
+            "name" : "validation",
+            "num_bytes" : 65875
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 682,
+            "num_bytes" : 65875
+         }
+      }
+   },
+   "eng-nds" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 357964,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 178982,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "validation"
+         },
+         "test" : {
+            "num_bytes" : 178982,
+            "num_examples" : 2500,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 5000,
+      "config_name" : "eng-nds",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "eng-jpn_Kana" : {
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_bytes" : 51632,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 503,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 51632,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 503,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 103264,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1006,
+      "config_name" : "eng-jpn_Kana",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "heb-tur" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 2754,
+      "config_name" : "heb-tur",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1377,
+            "name" : "validation",
+            "num_bytes" : 104779
+         },
+         "test" : {
+            "num_bytes" : 104779,
+            "name" : "test",
+            "num_examples" : 1377,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 209558
+   },
+   "tur-cmn_Hant" : {
+      "dataset_size" : 396,
+      "config_name" : "tur-cmn_Hant",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 26193,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 26193,
+            "name" : "test",
+            "num_examples" : 396,
+            "dataset_name" : "tatoeba_mt"
+         }
+      }
+   },
+   "fin-spa" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 5026,
+      "config_name" : "fin-spa",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 186328,
+            "num_examples" : 2513,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2513,
+            "num_bytes" : 186328
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 372656,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "crh-tur" : {
+      "splits" : {
+         "test" : {
+            "num_bytes" : 14520,
+            "name" : "test",
+            "num_examples" : 208,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 14520,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 208,
+      "config_name" : "crh-tur"
+   },
+   "bel-zho" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 29413,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 325,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 29413
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 325,
+      "config_name" : "bel-zho",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "epo-vol" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "epo-vol",
+      "dataset_size" : 1638,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 110112,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 55056,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 819
+         },
+         "test" : {
+            "num_bytes" : 55056,
+            "num_examples" : 819,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "ber_Latn-epo" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 133008,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 1575,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 133008
+         }
+      },
+      "config_name" : "ber_Latn-epo",
+      "dataset_size" : 1575,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "ina-spa" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "ina-spa",
+      "dataset_size" : 3226,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1613,
+            "name" : "test",
+            "num_bytes" : 165605
+         },
+         "validation" : {
+            "num_bytes" : 165605,
+            "name" : "validation",
+            "num_examples" : 1613,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 331210,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "epo-hbs" : {
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500,
+            "num_bytes" : 258348
+         },
+         "validation" : {
+            "num_bytes" : 258348,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "validation"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 516696,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 5000,
+      "config_name" : "epo-hbs",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "bel-deu" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 551,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 56902
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 56902,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 551,
+      "config_name" : "bel-deu"
+   },
+   "ita-jpn_Hira" : {
+      "size_in_bytes" : 483068,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 241534,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2326
+         },
+         "validation" : {
+            "num_bytes" : 241534,
+            "name" : "validation",
+            "num_examples" : 2326,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ita-jpn_Hira",
+      "dataset_size" : 4652,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "dan-jpn" : {
+      "dataset_size" : 904,
+      "config_name" : "dan-jpn",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 74725,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 74725,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 904,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null
+   },
+   "msa-msa" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 81386,
+            "num_examples" : 958,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 81386,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 958,
+      "config_name" : "msa-msa",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "uig_Arab-cmn_Hans" : {
+      "size_in_bytes" : 151405,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 151405,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1383
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1383,
+      "config_name" : "uig_Arab-cmn_Hans",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "eng-bos_Latn" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 19082,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 301
+         },
+         "validation" : {
+            "num_bytes" : 19082,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 301,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 38164,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 602,
+      "config_name" : "eng-bos_Latn",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-ukr" : {
+      "size_in_bytes" : 2331518,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 1165759,
+            "name" : "validation",
+            "num_examples" : 13127,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 1165759,
+            "num_examples" : 13127,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 26254,
+      "config_name" : "eng-ukr",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "ina-yid" : {
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 788,
+            "name" : "validation",
+            "num_bytes" : 73509
+         },
+         "test" : {
+            "num_bytes" : 73509,
+            "name" : "test",
+            "num_examples" : 788,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 147018,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1576,
+      "config_name" : "ina-yid"
+   },
+   "ita-nld" : {
+      "config_name" : "ita-nld",
+      "dataset_size" : 5156,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 346396,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2578,
+            "name" : "validation",
+            "num_bytes" : 173198
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2578,
+            "num_bytes" : 173198
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "rus-rus" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 5000,
+      "config_name" : "rus-rus",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 312967,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2500,
+            "name" : "validation"
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 2500,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 312967
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 625934
+   },
+   "kab-rus" : {
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 423,
+            "num_bytes" : 39530
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 39530,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "kab-rus",
+      "dataset_size" : 423,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "fra-ido" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 20748,
+            "name" : "test",
+            "num_examples" : 296,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 20748,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 296
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 41496,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 592,
+      "config_name" : "fra-ido",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "cat-ukr" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "cat-ukr",
+      "dataset_size" : 456,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 35263,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 456,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 35263
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "nor-pol" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "nor-pol",
+      "dataset_size" : 281,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 21984,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 21984,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 281,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "eng-glg" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 1015,
+      "config_name" : "eng-glg",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 85016,
+            "name" : "test",
+            "num_examples" : 1015,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 85016,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "deu-jpn_Hani" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 129301,
+            "num_examples" : 1290,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 129301,
+            "name" : "validation",
+            "num_examples" : 1290,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 258602,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 2580,
+      "config_name" : "deu-jpn_Hani",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "srp_Latn-ukr" : {
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 348,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 25013
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 25013,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 348,
+      "config_name" : "srp_Latn-ukr",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "fin-kor" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "fin-kor",
+      "dataset_size" : 422,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 422,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 33898
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 33898,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "deu-eus" : {
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 456,
+            "name" : "test",
+            "num_bytes" : 35688
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 35688,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 456,
+      "config_name" : "deu-eus",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "nld-nld" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "nld-nld",
+      "dataset_size" : 2000,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 74782,
+            "num_examples" : 1000,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1000,
+            "name" : "validation",
+            "num_bytes" : 74782
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 149564,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "deu-pol" : {
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "size_in_bytes" : 810894,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 405447,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 5000,
+            "name" : "test"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 5000,
+            "num_bytes" : 405447
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "deu-pol",
+      "dataset_size" : 10000,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "deu-lit" : {
+      "dataset_size" : 2230,
+      "config_name" : "deu-lit",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 185060,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 1115,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 92530
+         },
+         "test" : {
+            "num_bytes" : 92530,
+            "name" : "test",
+            "num_examples" : 1115,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null
+   },
+   "rus-cmn_Hans" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 116284,
+            "name" : "validation",
+            "num_examples" : 1086,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1086,
+            "num_bytes" : 116284
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 232568,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 2172,
+      "config_name" : "rus-cmn_Hans",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-nov" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "eng-nov",
+      "dataset_size" : 217,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 14871,
+            "num_examples" : 217,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 14871,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "deu-lfn_Latn" : {
+      "dataset_size" : 582,
+      "config_name" : "deu-lfn_Latn",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 51672,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 25836,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 291,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 25836,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 291,
+            "name" : "validation"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "jpn_Kana-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 508,
+      "config_name" : "jpn_Kana-spa",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 49154,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 254,
+            "name" : "test",
+            "num_bytes" : 24577
+         },
+         "validation" : {
+            "num_bytes" : 24577,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 254,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "deu-toki" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 20228,
+      "config_name" : "deu-toki",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 945160,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10114
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 10114,
+            "num_bytes" : 945160
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1890320,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null
+   },
+   "grn-por" : {
+      "config_name" : "grn-por",
+      "dataset_size" : 448,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "size_in_bytes" : 35282,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_examples" : 224,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 17641
+         },
+         "validation" : {
+            "num_examples" : 224,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 17641
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "deu-swe" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "splits" : {
+         "test" : {
+            "num_examples" : 3410,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 254512
+         },
+         "validation" : {
+            "num_bytes" : 254512,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 3410,
+            "name" : "validation"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 509024,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 6820,
+      "config_name" : "deu-swe",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-nor" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-nor",
+      "dataset_size" : 10000,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 810614,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 5000,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 405307
+         },
+         "validation" : {
+            "num_examples" : 5000,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 405307
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      }
+   },
+   "isl-jpn_Hira" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 18260,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 211,
+            "num_bytes" : 18260
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "isl-jpn_Hira",
+      "dataset_size" : 211,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "ido-lfn" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 696,
+      "config_name" : "ido-lfn",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 23776,
+            "name" : "validation",
+            "num_examples" : 348,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 23776,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 348
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 47552
+   },
+   "nor-swe" : {
+      "dataset_size" : 566,
+      "config_name" : "nor-swe",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 39044,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 566,
+            "name" : "test",
+            "num_bytes" : 39044
+         }
+      },
+      "task_templates" : null
+   },
+   "ell-swe" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 253,
+      "config_name" : "ell-swe",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 253,
+            "name" : "test",
+            "num_bytes" : 21696
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 21696
+   },
+   "bul-ukr" : {
+      "config_name" : "bul-ukr",
+      "dataset_size" : 1020,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 88243,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 88243,
+            "name" : "test",
+            "num_examples" : 1020,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "eng-nno" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 920,
+      "config_name" : "eng-nno",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 34843,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 460
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 460,
+            "name" : "validation",
+            "num_bytes" : 34843
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 69686
+   },
+   "eng-hun" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 26074,
+      "config_name" : "eng-hun",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 1918548,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 13037,
+            "name" : "test",
+            "num_bytes" : 959274
+         },
+         "validation" : {
+            "num_bytes" : 959274,
+            "name" : "validation",
+            "num_examples" : 13037,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "tur-zho" : {
+      "size_in_bytes" : 67341,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 986,
+            "num_bytes" : 67341
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 986,
+      "config_name" : "tur-zho",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "jpn_Hani-nld" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "jpn_Hani-nld",
+      "dataset_size" : 852,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 70606,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 35303,
+            "num_examples" : 426,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 426,
+            "num_bytes" : 35303
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "deu-ltz" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 347,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 23796
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 23796,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "deu-ltz",
+      "dataset_size" : 347
+   },
+   "ara-ber" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ara-ber",
+      "dataset_size" : 4634,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 419756,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 2317,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 209878
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2317,
+            "num_bytes" : 209878
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "eng-hrv" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 208844,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 104422,
+            "name" : "test",
+            "num_examples" : 1480,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 104422,
+            "name" : "validation",
+            "num_examples" : 1480,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "config_name" : "eng-hrv",
+      "dataset_size" : 2960,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "lad-spa" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 34666,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 276,
+            "name" : "validation",
+            "num_bytes" : 17333
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 276,
+            "name" : "test",
+            "num_bytes" : 17333
+         }
+      },
+      "config_name" : "lad-spa",
+      "dataset_size" : 552,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "deu-tlh" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 156376,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 78188,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 1099
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 1099,
+            "name" : "test",
+            "num_bytes" : 78188
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 2198,
+      "config_name" : "deu-tlh",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "fra-kor" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "fra-kor",
+      "dataset_size" : 321,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 321,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 32027
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 32027,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "jpn-kor" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 622,
+      "config_name" : "jpn-kor",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 622,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 64304
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 64304,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "epo-tur" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 10000,
+      "config_name" : "epo-tur",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 719346,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 359673,
+            "name" : "validation",
+            "num_examples" : 5000,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 5000,
+            "name" : "test",
+            "num_bytes" : 359673
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "eng-pms" : {
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 21102,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 269,
+            "num_bytes" : 21102
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "eng-pms",
+      "dataset_size" : 269,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "jpn_Hira-mar" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 254,
+      "config_name" : "jpn_Hira-mar",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 254,
+            "name" : "test",
+            "num_bytes" : 31703
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 31703,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      }
+   },
+   "lat-yid" : {
+      "config_name" : "lat-yid",
+      "dataset_size" : 916,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "size_in_bytes" : 82624,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 41312,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 458
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 458,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 41312
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "deu-gos" : {
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 207,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 12260
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 12260,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 207,
+      "config_name" : "deu-gos",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "lad_Latn-spa" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 14299,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 239,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 14299,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "lad_Latn-spa",
+      "dataset_size" : 239,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "fra-ron" : {
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 140669,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 140669,
+            "name" : "test",
+            "num_examples" : 1925,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 1925,
+      "config_name" : "fra-ron",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "bel-epo" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 75827,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 75827,
+            "num_examples" : 735,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 735,
+      "config_name" : "bel-epo",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "bua_Cyrl-rus" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 74416,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 805
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 74416,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "bua_Cyrl-rus",
+      "dataset_size" : 805,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "srp_Cyrl-rus" : {
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 881,
+            "name" : "validation",
+            "num_bytes" : 98682
+         },
+         "test" : {
+            "num_bytes" : 98682,
+            "num_examples" : 881,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 197364,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "srp_Cyrl-rus",
+      "dataset_size" : 1762,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ara-tur" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 218534,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 1263,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 109267
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1263,
+            "num_bytes" : 109267
+         }
+      },
+      "config_name" : "ara-tur",
+      "dataset_size" : 2526,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "lat-spa" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 6258,
+      "config_name" : "lat-spa",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 3129,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 346603
+         },
+         "validation" : {
+            "num_bytes" : 346603,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 3129
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 693206,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      }
+   },
+   "kab-kab" : {
+      "dataset_size" : 1986,
+      "config_name" : "kab-kab",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 185664,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 92832,
+            "name" : "test",
+            "num_examples" : 993,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "num_bytes" : 92832,
+            "name" : "validation",
+            "num_examples" : 993,
+            "dataset_name" : "tatoeba_mt"
+         }
+      }
+   },
+   "deu-jpn" : {
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 2467024,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 1233512,
+            "name" : "validation",
+            "num_examples" : 11428,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 1233512,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 11428
+         }
+      },
+      "task_templates" : null,
+      "dataset_size" : 22856,
+      "config_name" : "deu-jpn",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "jpn_Hira-spa" : {
+      "size_in_bytes" : 1876100,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 9379,
+            "num_bytes" : 938050
+         },
+         "test" : {
+            "num_bytes" : 938050,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 9379,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 18758,
+      "config_name" : "jpn_Hira-spa",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "lad-yid" : {
+      "config_name" : "lad-yid",
+      "dataset_size" : 1208,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "size_in_bytes" : 110266,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 604,
+            "name" : "validation",
+            "num_bytes" : 55133
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 604,
+            "name" : "test",
+            "num_bytes" : 55133
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "ara-eng" : {
+      "dataset_size" : 20610,
+      "config_name" : "ara-eng",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 1876342,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_examples" : 10305,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 938171
+         },
+         "test" : {
+            "num_bytes" : 938171,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10305,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null
+   },
+   "ber_Latn-fra" : {
+      "size_in_bytes" : 1515566,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10140,
+            "num_bytes" : 757783
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10140,
+            "name" : "test",
+            "num_bytes" : 757783
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ber_Latn-fra",
+      "dataset_size" : 20280,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "eng-fin" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "eng-fin",
+      "dataset_size" : 21380,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 843346,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10690
+         },
+         "test" : {
+            "num_examples" : 10690,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 843346
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1686692,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      }
+   },
+   "deu-lad" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "deu-lad",
+      "dataset_size" : 440,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 220,
+            "num_bytes" : 14517
+         },
+         "test" : {
+            "num_bytes" : 14517,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 220
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 29034,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      }
+   },
+   "dan-fra" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "dan-fra",
+      "dataset_size" : 3462,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 120949,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1731
+         },
+         "validation" : {
+            "num_bytes" : 120949,
+            "num_examples" : 1731,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 241898,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "kaz-rus" : {
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1400,
+            "num_bytes" : 337480
+         },
+         "validation" : {
+            "num_bytes" : 337480,
+            "name" : "validation",
+            "num_examples" : 1400,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 674960,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "kaz-rus",
+      "dataset_size" : 2800,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-jpn_Hira" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "eng-jpn_Hira",
+      "dataset_size" : 23662,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 1289993,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 11831,
+            "name" : "test"
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 11831,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1289993
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 2579986,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "rus-toki_Latn" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 899,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 86569
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 86569,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "rus-toki_Latn",
+      "dataset_size" : 899,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ido-spa" : {
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "size_in_bytes" : 37025,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 37025,
+            "num_examples" : 499,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 499,
+      "config_name" : "ido-spa",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "nld-nor" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 15232,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 203,
+            "num_bytes" : 15232
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 203,
+      "config_name" : "nld-nor",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-yue_Hans" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-yue_Hans",
+      "dataset_size" : 1354,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 127430,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 63715,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 677,
+            "name" : "validation"
+         },
+         "test" : {
+            "num_bytes" : 63715,
+            "num_examples" : 677,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "ido-yid" : {
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "size_in_bytes" : 75298,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 37649,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 426
+         },
+         "test" : {
+            "num_bytes" : 37649,
+            "name" : "test",
+            "num_examples" : 426,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "dataset_size" : 852,
+      "config_name" : "ido-yid",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "fra-cmn_Hant" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fra-cmn_Hant",
+      "dataset_size" : 4264,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 349704,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 2132,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 174852
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2132,
+            "num_bytes" : 174852
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "eng-nld" : {
+      "size_in_bytes" : 1873352,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 936676,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 12696
+         },
+         "validation" : {
+            "num_bytes" : 936676,
+            "name" : "validation",
+            "num_examples" : 12696,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 25392,
+      "config_name" : "eng-nld",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "dan-por" : {
+      "size_in_bytes" : 54721,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 873,
+            "name" : "test",
+            "num_bytes" : 54721
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "dan-por",
+      "dataset_size" : 873,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "ita-ukr" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 747740,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 373870,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 5000,
+            "name" : "test"
+         },
+         "validation" : {
+            "num_bytes" : 373870,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 5000,
+            "name" : "validation"
+         }
+      },
+      "config_name" : "ita-ukr",
+      "dataset_size" : 10000,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "glg-spa" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 4242,
+      "config_name" : "glg-spa",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 367802,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 183901,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2121,
+            "name" : "test"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2121,
+            "num_bytes" : 183901
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      }
+   },
+   "eng-uig_Arab" : {
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 314408,
+            "name" : "validation",
+            "num_examples" : 3021,
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 3021,
+            "num_bytes" : 314408
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 628816,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "eng-uig_Arab",
+      "dataset_size" : 6042
+   },
+   "deu-rus" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 12800,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 1345955
+         },
+         "validation" : {
+            "num_bytes" : 1345955,
+            "name" : "validation",
+            "num_examples" : 12800,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "task_templates" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 2691910,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "deu-rus",
+      "dataset_size" : 25600
+   },
+   "ces-fra" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 438,
+      "config_name" : "ces-fra",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 34150,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 34150,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 438
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      }
+   },
+   "hin-urd" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 240,
+      "config_name" : "hin-urd",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 33691,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 240,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 33691
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      }
+   },
+   "nno-nob" : {
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 467,
+            "num_bytes" : 31927
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 31927,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "nno-nob",
+      "dataset_size" : 467,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-pam" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "eng-pam",
+      "dataset_size" : 2000,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "size_in_bytes" : 150082,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 75041,
+            "num_examples" : 1000,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 1000,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 75041
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "patch" : 0,
+         "major" : 0
+      }
+   },
+   "por-ukr" : {
+      "dataset_size" : 6744,
+      "config_name" : "por-ukr",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 570034,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 285017,
+            "num_examples" : 3372,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "test" : {
+            "num_bytes" : 285017,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 3372,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null
+   },
+   "jpn_Hira-vie" : {
+      "dataset_size" : 1856,
+      "config_name" : "jpn_Hira-vie",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07"
+      },
+      "size_in_bytes" : 214416,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 107208,
+            "num_examples" : 928,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 928,
+            "name" : "validation",
+            "num_bytes" : 107208
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "ell-rus" : {
+      "size_in_bytes" : 565016,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 282508,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 2500
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 2500,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 282508
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 5000,
+      "config_name" : "ell-rus",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "hrv-spa" : {
+      "config_name" : "hrv-spa",
+      "dataset_size" : 254,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 16960,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 254,
+            "num_bytes" : 16960
+         }
+      }
+   },
+   "nor-rus" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 171086,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 1279,
+            "num_bytes" : 171086
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1279,
+      "config_name" : "nor-rus",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "eng-ota_Latn" : {
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 317,
+            "name" : "test",
+            "num_bytes" : 24982
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 24982,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "eng-ota_Latn",
+      "dataset_size" : 317,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "tat-vie" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 284,
+      "config_name" : "tat-vie",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 30530,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 284,
+            "num_bytes" : 30530
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      }
+   },
+   "eng-jav" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 262,
+      "config_name" : "eng-jav",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 19245,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 262,
+            "num_bytes" : 19245
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      }
+   },
+   "cmn_Hans-wuu" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 147728,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 815,
+            "num_bytes" : 73864
+         },
+         "validation" : {
+            "num_bytes" : 73864,
+            "name" : "validation",
+            "num_examples" : 815,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "config_name" : "cmn_Hans-wuu",
+      "dataset_size" : 1630,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "jpn_Hira-pol" : {
+      "size_in_bytes" : 1849862,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "splits" : {
+         "test" : {
+            "num_examples" : 8765,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 924931
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 8765,
+            "name" : "validation",
+            "num_bytes" : 924931
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "jpn_Hira-pol",
+      "dataset_size" : 17530,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null
+   },
+   "lat-lfn" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 348,
+            "name" : "test",
+            "num_bytes" : 22981
+         },
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 348,
+            "name" : "validation",
+            "num_bytes" : 22981
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 45962,
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 696,
+      "config_name" : "lat-lfn"
+   },
+   "epo-ina" : {
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 2644,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 250841
+         },
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2644,
+            "name" : "test",
+            "num_bytes" : 250841
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 501682,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "epo-ina",
+      "dataset_size" : 5288,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-lat" : {
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 10298,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 968004
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 10298,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 968004
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 1936008,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 20596,
+      "config_name" : "eng-lat",
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "eng-ota_Arab" : {
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 33731,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 371,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 33731
+         }
+      },
+      "task_templates" : null,
+      "config_name" : "eng-ota_Arab",
+      "dataset_size" : 371,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "epo-ita" : {
+      "config_name" : "epo-ita",
+      "dataset_size" : 20606,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "features" : {
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "size_in_bytes" : 1530536,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_examples" : 10303,
+            "name" : "validation",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 765268
+         },
+         "test" : {
+            "num_bytes" : 765268,
+            "name" : "test",
+            "num_examples" : 10303,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using."
+   },
+   "fra-tgl" : {
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 69627,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_examples" : 836,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 69627
+         }
+      },
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "fra-tgl",
+      "dataset_size" : 836,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "lad-lfn" : {
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 23973,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 335,
+            "name" : "validation"
+         },
+         "test" : {
+            "num_bytes" : 23973,
+            "name" : "test",
+            "num_examples" : 335,
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 47946,
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "lad-lfn",
+      "dataset_size" : 670
+   },
+   "deu-ile" : {
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "deu-ile",
+      "dataset_size" : 4006,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "splits" : {
+         "validation" : {
+            "num_bytes" : 137021,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 2003
+         },
+         "test" : {
+            "name" : "test",
+            "num_examples" : 2003,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 137021
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 274042,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}"
+   },
+   "chm-rus" : {
+      "splits" : {
+         "validation" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2750,
+            "name" : "validation",
+            "num_bytes" : 381966
+         },
+         "test" : {
+            "num_bytes" : 381966,
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 2750,
+            "name" : "test"
+         }
+      },
+      "task_templates" : null,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 763932,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null
+      },
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "dataset_size" : 5500,
+      "config_name" : "chm-rus"
+   },
+   "eng-que" : {
+      "dataset_size" : 257,
+      "config_name" : "eng-que",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "description" : null,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 15806,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 257,
+            "name" : "test",
+            "num_bytes" : 15806
+         }
+      }
+   },
+   "ita-nds" : {
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "config_name" : "ita-nds",
+      "dataset_size" : 313,
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 21384,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "name" : "test",
+            "num_examples" : 313,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 21384
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "major" : 0,
+         "patch" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "spa-tur" : {
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "dataset_size" : 21230,
+      "config_name" : "spa-tur",
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 10615,
+            "name" : "test",
+            "num_bytes" : 743546
+         },
+         "validation" : {
+            "num_bytes" : 743546,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "validation",
+            "num_examples" : 10615
+         }
+      },
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "size_in_bytes" : 1487092,
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "targetString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      }
+   },
+   "zho-zho" : {
+      "features" : {
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetlang" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "major" : 0,
+         "patch" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "size_in_bytes" : 528672,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "supervised_keys" : null,
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 264336,
+            "num_examples" : 3114,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         },
+         "validation" : {
+            "name" : "validation",
+            "num_examples" : 3114,
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 264336
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "config_name" : "zho-zho",
+      "dataset_size" : 6228,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt"
+   },
+   "jbo_Latn-jpn_Hira" : {
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "config_name" : "jbo_Latn-jpn_Hira",
+      "dataset_size" : 826,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0",
+      "version" : {
+         "description" : null,
+         "version_str" : "v2021-08-07",
+         "minor" : 0,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "dataset_name" : "tatoeba_mt",
+            "num_examples" : 826,
+            "name" : "test",
+            "num_bytes" : 72892
+         }
+      },
+      "task_templates" : null,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 72892
+   },
+   "eng-tat" : {
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 143755,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_examples" : 1451,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt",
+            "num_bytes" : 143755
+         }
+      },
+      "task_templates" : null,
+      "version" : {
+         "patch" : 0,
+         "major" : 0,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null
+      },
+      "features" : {
+         "sourceString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         }
+      },
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "dataset_size" : 1451,
+      "config_name" : "eng-tat",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/"
+   },
+   "vie-cmn_Hans" : {
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetString" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         },
+         "sourceLang" : {
+            "dtype" : "string",
+            "id" : null,
+            "_type" : "Value"
+         }
+      },
+      "version" : {
+         "description" : null,
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "major" : 0,
+         "patch" : 0
+      },
+      "task_templates" : null,
+      "splits" : {
+         "test" : {
+            "num_bytes" : 38707,
+            "num_examples" : 345,
+            "name" : "test",
+            "dataset_name" : "tatoeba_mt"
+         }
+      },
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "size_in_bytes" : 38707,
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "post_processed" : null,
+      "config_name" : "vie-cmn_Hans",
+      "dataset_size" : 345,
+      "builder_name" : "tatoeba_mt",
+      "license" : "cc-by-2.0"
+   },
+   "ara-ita" : {
+      "dataset_size" : 235,
+      "config_name" : "ara-ita",
+      "post_processed" : null,
+      "homepage" : "https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+      "license" : "cc-by-2.0",
+      "builder_name" : "tatoeba_mt",
+      "version" : {
+         "minor" : 0,
+         "version_str" : "v2021-08-07",
+         "description" : null,
+         "patch" : 0,
+         "major" : 0
+      },
+      "features" : {
+         "sourceString" : {
+            "id" : null,
+            "dtype" : "string",
+            "_type" : "Value"
+         },
+         "targetlang" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "targetString" : {
+            "_type" : "Value",
+            "dtype" : "string",
+            "id" : null
+         },
+         "sourceLang" : {
+            "_type" : "Value",
+            "id" : null,
+            "dtype" : "string"
+         }
+      },
+      "supervised_keys" : null,
+      "citation" : "@inproceedings{tiedemann-2020-tatoeba,\n    title = \"The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}\",\n    author = {Tiedemann, J{\\\"o}rg},\n    booktitle = \"Proceedings of the Fifth Conference on Machine Translation\",\n    month = nov,\n    year = \"2020\",\n    address = \"Online\",\n    publisher = \"Association for Computational Linguistics\",\n    url = \"https://aclanthology.org/2020.wmt-1.139\",\n    pages = \"1174--1182\",\n}",
+      "size_in_bytes" : 18780,
+      "description" : "The Tatoeba Translation Challenge is a multilingual data set of machine translation benchmarks derived from user-contributed translations collected by [Tatoeba.org](https://tatoeba.org/) and provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This dataset includes test and development data sorted by language pair. It includes test sets for hundreds of language pairs and is continuously updated. Please, check the version number tag to refer to the release that your are using.",
+      "splits" : {
+         "test" : {
+            "num_bytes" : 18780,
+            "dataset_name" : "tatoeba_mt",
+            "name" : "test",
+            "num_examples" : 235
+         }
+      },
+      "task_templates" : null
+   }
+}
diff --git a/dev/tatoeba-dev.afb-eng.tsv b/dev/tatoeba-dev.afb-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..05b2f2eb332cb040931f1b340b91d4c0bb2da661
--- /dev/null
+++ b/dev/tatoeba-dev.afb-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30b6448e3a489c4bddc53c5354048015740a4d014d0b1c3f5f19a9eb44f3c4a1
+size 2734
diff --git a/dev/tatoeba-dev.afr-eng.tsv b/dev/tatoeba-dev.afr-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..17b49e4e60f41aed5977e4eb10d084e733715ae3
--- /dev/null
+++ b/dev/tatoeba-dev.afr-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09bbe49b537146f79de92ee263ecc5cb198bfa50a6095224335cbe39b8410aa9
+size 69888
diff --git a/dev/tatoeba-dev.ara-ber.tsv b/dev/tatoeba-dev.ara-ber.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1a3737318b2277da0197d921a4f61566e2a34cda
--- /dev/null
+++ b/dev/tatoeba-dev.ara-ber.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3226fd66a4da99ae3ba631b43210656b3fab959ecedafb25ac3c260047627359
+size 95068
diff --git a/dev/tatoeba-dev.ara-ber_Latn.tsv b/dev/tatoeba-dev.ara-ber_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..42dad8c62dbf789dd91764ab01fe7dc91f098432
--- /dev/null
+++ b/dev/tatoeba-dev.ara-ber_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ba22347e48fe4397640cf22bfac7c3e35219c8f673ee371e6be71c95d298c07
+size 94416
diff --git a/dev/tatoeba-dev.ara-deu.tsv b/dev/tatoeba-dev.ara-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9ac801eddf9f17a34c981038953b94c03a88ffdc
--- /dev/null
+++ b/dev/tatoeba-dev.ara-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f3e578c1a246df710efa36ea4dd5024658c61fb6a6486220e9aef050be0e411
+size 87602
diff --git a/dev/tatoeba-dev.ara-eng.tsv b/dev/tatoeba-dev.ara-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c4c81b35bf3bebd5107d1dc5ac35e7c458268295
--- /dev/null
+++ b/dev/tatoeba-dev.ara-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82dc113098ecd50645405326a2c8eff70d07328e6a4d85d78e5e004a12dc1d65
+size 1778245
diff --git a/dev/tatoeba-dev.ara-epo.tsv b/dev/tatoeba-dev.ara-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..64a8a0cce6f9fec92cac51aef60a14a5effaadb9
--- /dev/null
+++ b/dev/tatoeba-dev.ara-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66cbd2635506b8a8e6d86976077cbba95c6a8c06c92e9bcb1d98be262ed49a7b
+size 7818
diff --git a/dev/tatoeba-dev.ara-fra.tsv b/dev/tatoeba-dev.ara-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..14879dc948d36c91a1fb2a03415e7f7e450265f8
--- /dev/null
+++ b/dev/tatoeba-dev.ara-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a46a2be4f406113c27d20782280b8ac6008fd4be863e5141d2b24dcba3c71bfb
+size 3669
diff --git a/dev/tatoeba-dev.ara-heb.tsv b/dev/tatoeba-dev.ara-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..73ef9cf9317eae39b55d09bbe0fe53b6a0d1b475
--- /dev/null
+++ b/dev/tatoeba-dev.ara-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9c1d7964b24db80038196bc66c2d96040caf75d2cc90a65f32bd861c52b6e58
+size 8924
diff --git a/dev/tatoeba-dev.ara-rus.tsv b/dev/tatoeba-dev.ara-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bd36dcec7db25b58d2555108397468b9bd2ba563
--- /dev/null
+++ b/dev/tatoeba-dev.ara-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a888e195573dfe6def6199e4fffa65fc72295fd4ab3fee72650cf2056c711c68
+size 122685
diff --git a/dev/tatoeba-dev.ara-spa.tsv b/dev/tatoeba-dev.ara-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..93340a789d36ad1edf47c1a4eee078d32febbe1d
--- /dev/null
+++ b/dev/tatoeba-dev.ara-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4cce607b083f134ea075be9506e95db4adbb4c5d1f2dd71c08a6b5888d3c518b
+size 81989
diff --git a/dev/tatoeba-dev.ara-tur.tsv b/dev/tatoeba-dev.ara-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f76be9462453300e60eeb3e98d8e3e03a5c6eae5
--- /dev/null
+++ b/dev/tatoeba-dev.ara-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab145cda61aca57e64c75725f0b3a2a73adc031428c149434e515d30ca7a57fb
+size 4914
diff --git a/dev/tatoeba-dev.arq-eng.tsv b/dev/tatoeba-dev.arq-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1ff07e1e4e7b3e084343473fb19838c2af806084
--- /dev/null
+++ b/dev/tatoeba-dev.arq-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:655b1495cd2264b0c807c3b1fb0a885cd807e067de229ac547f8e70462bcb13d
+size 68363
diff --git a/dev/tatoeba-dev.ary-eng.tsv b/dev/tatoeba-dev.ary-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d802f8fca522035f646847d92afb1a56e050fa4f
--- /dev/null
+++ b/dev/tatoeba-dev.ary-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0e0496972f0998da1e607e263ba6e4bfafca79dfefa8aaa4d3134b1109f6313
+size 1877
diff --git a/dev/tatoeba-dev.arz-eng.tsv b/dev/tatoeba-dev.arz-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e745f0dbd621ef044b298d30b94d7c1f9cd988cd
--- /dev/null
+++ b/dev/tatoeba-dev.arz-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6147b65cef1ce6b9886210b15ef47d83c7c3bd50db3c80d3889a5341f543229a
+size 32064
diff --git a/dev/tatoeba-dev.avk-fra.tsv b/dev/tatoeba-dev.avk-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..03e7372340e99fa3a0d9a7ecb9e655a7d4103fae
--- /dev/null
+++ b/dev/tatoeba-dev.avk-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3e889a0d5f71e75e47331be119abaa56b4c1c4ab836c543310e7115acc05da1
+size 3230
diff --git a/dev/tatoeba-dev.aze-eng.tsv b/dev/tatoeba-dev.aze-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..59305cc6833a05ccecf3204d52551f25143e17a5
--- /dev/null
+++ b/dev/tatoeba-dev.aze-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0280e17dedb307d966484ebd190db5673b0ce3e8422aad8946d8d0018fb5a996
+size 71379
diff --git a/dev/tatoeba-dev.bel-eng.tsv b/dev/tatoeba-dev.bel-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..42d0aa422e0cc3929b48b5df2f968d19cf1f3e27
--- /dev/null
+++ b/dev/tatoeba-dev.bel-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ab514f185f40be210f00cc65823ccc8c9e9e0ed8dfa2875c0e895b8121d4766
+size 420192
diff --git a/dev/tatoeba-dev.bel-rus.tsv b/dev/tatoeba-dev.bel-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f103dad464ba7ff0d6ca3ec78c07e1fe21a69868
--- /dev/null
+++ b/dev/tatoeba-dev.bel-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ea63e9eec7728edcbb559ee32147848d0547a44512dc8bbdfb6e3c2b7080e27
+size 404286
diff --git a/dev/tatoeba-dev.bel-ukr.tsv b/dev/tatoeba-dev.bel-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..81793edc5c68bdc10952b13d9232c58ff05d2c6d
--- /dev/null
+++ b/dev/tatoeba-dev.bel-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c71cd530d088a881f3bb5074946ed403b369a64de96d38b02d0e0abe5537c13b
+size 117733
diff --git a/dev/tatoeba-dev.ben-eng.tsv b/dev/tatoeba-dev.ben-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a72971d7cf13e07bfd8cdf73cd651546a5cece5b
--- /dev/null
+++ b/dev/tatoeba-dev.ben-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4f42431f26c46497d05b57c6cd4be46981336bb5185eef56f2990eb5337759f
+size 248865
diff --git a/dev/tatoeba-dev.ber-eng.tsv b/dev/tatoeba-dev.ber-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3abed0523bef8c3a36723d57c8d9006a73308351
--- /dev/null
+++ b/dev/tatoeba-dev.ber-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fb438387270a302c685816cff07f94dc85f2b350a0e3f33bc7bfe96647d8de3
+size 8894631
diff --git a/dev/tatoeba-dev.ber-fra.tsv b/dev/tatoeba-dev.ber-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e6e5869c4c3f3317e95e1f25e47aec2e9dc98535
--- /dev/null
+++ b/dev/tatoeba-dev.ber-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:640485f0f59b64177209dcbac7813a4d63816bb7915de86ff2cff40d58dbe262
+size 2588616
diff --git a/dev/tatoeba-dev.ber-spa.tsv b/dev/tatoeba-dev.ber-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..efc9fe0116171dc3baccba92bda1ca8bf1ec56d3
--- /dev/null
+++ b/dev/tatoeba-dev.ber-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18b2568372aff944524006649e8beed4075082883b9801e67b1c330cadc4e248
+size 1130816
diff --git a/dev/tatoeba-dev.ber_Latn-eng.tsv b/dev/tatoeba-dev.ber_Latn-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..aa2926993f9e5f9e3a336f112e4f2c850e71492f
--- /dev/null
+++ b/dev/tatoeba-dev.ber_Latn-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a2115392ced2e699996dabbef1bcc322453c10888a843baa864e94cb5667644
+size 8892582
diff --git a/dev/tatoeba-dev.ber_Latn-fra.tsv b/dev/tatoeba-dev.ber_Latn-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a03563f93699a7bf181d5024bad8a7bfb448c951
--- /dev/null
+++ b/dev/tatoeba-dev.ber_Latn-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea2cb33500094e3657c6832aad5fe1827c05cf86755e18d03209464cc624c4c5
+size 2588010
diff --git a/dev/tatoeba-dev.ber_Latn-spa.tsv b/dev/tatoeba-dev.ber_Latn-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d31df2c5c4d0e94648786f27d6e2d5ef1df4e1fb
--- /dev/null
+++ b/dev/tatoeba-dev.ber_Latn-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52ad6e388ea266313e67e959326bcc202ed04d345819879e06fdabcc1b7d9ef4
+size 1130652
diff --git a/dev/tatoeba-dev.bos_Latn-rus.tsv b/dev/tatoeba-dev.bos_Latn-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5d94262c17ab729f31a4b91879f2aee12e11220f
--- /dev/null
+++ b/dev/tatoeba-dev.bos_Latn-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:354141531acb58e7036b24c62c442f577dc02d83086d7658ebd220fe801de23a
+size 2414
diff --git a/dev/tatoeba-dev.bre-fra.tsv b/dev/tatoeba-dev.bre-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fe17b159ed314f9c396074b96534cd41afab127f
--- /dev/null
+++ b/dev/tatoeba-dev.bre-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0da61aeb81ad1c51302b6a7fde78aa72d3c1f400f24a02382b5c64962b402f5d
+size 182936
diff --git a/dev/tatoeba-dev.bul-eng.tsv b/dev/tatoeba-dev.bul-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cd9f39172a952a52df6e57072caa1193af298f44
--- /dev/null
+++ b/dev/tatoeba-dev.bul-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4071be08ac440eeea715993d6dd1a6fee3624fcf46d7dc9fe2378c99ff8c225
+size 745796
diff --git a/dev/tatoeba-dev.bul-ita.tsv b/dev/tatoeba-dev.bul-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d4312aa396bbfd6ad34725480b9ba2c428296ed1
--- /dev/null
+++ b/dev/tatoeba-dev.bul-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16c5b9b0e1b562f7a4f39c6613f6bd9f0b53af9ae4b28840929ed48a67bb61a6
+size 389298
diff --git a/dev/tatoeba-dev.bul-rus.tsv b/dev/tatoeba-dev.bul-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..38854755b9e2a8d693c886bf7f0b4468bcd95786
--- /dev/null
+++ b/dev/tatoeba-dev.bul-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:938f7b3d7bd3bb44fb76d0d99f4a78be164bdff7382b23dc7587fbca76321e19
+size 112065
diff --git a/dev/tatoeba-dev.cat-eng.tsv b/dev/tatoeba-dev.cat-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ee8f93c374d27bc38d74fb6352806aaad0bf709f
--- /dev/null
+++ b/dev/tatoeba-dev.cat-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a360df54edb583f816e0f9f81befff17702b3131737b751f39fbb6a2d3583f9
+size 23465
diff --git a/dev/tatoeba-dev.cat-spa.tsv b/dev/tatoeba-dev.cat-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e3fbe16e76b9ff7c59a6cc917b421e52829834cb
--- /dev/null
+++ b/dev/tatoeba-dev.cat-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d6e66f10f75651deba728c288144e42b0549a400ff939962d3870634784a81e
+size 114582
diff --git a/dev/tatoeba-dev.cbk-eng.tsv b/dev/tatoeba-dev.cbk-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..375a3329358bb3b6fe0a3a68c5aaf6d8acb7d989
--- /dev/null
+++ b/dev/tatoeba-dev.cbk-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c543972563b87e7c7eb71273e5d10df717dbd87488f64f59cbb2d56341ae6115
+size 73581
diff --git a/dev/tatoeba-dev.ces-deu.tsv b/dev/tatoeba-dev.ces-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b49aa91e8f17875c04d9a4e9d43bc5a5cf3c4def
--- /dev/null
+++ b/dev/tatoeba-dev.ces-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dca1a2ae119ee7952d5502e64d309412d978fb19a6866b2b7014a06e92c547e4
+size 93678
diff --git a/dev/tatoeba-dev.ces-eng.tsv b/dev/tatoeba-dev.ces-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..50a7091a0d7d32d881819b27761af69d190b8c55
--- /dev/null
+++ b/dev/tatoeba-dev.ces-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d8ee65f9e190d6d572315a6571313da628c5a811c6b04b2d08d119a628b9553
+size 1212451
diff --git a/dev/tatoeba-dev.ces-epo.tsv b/dev/tatoeba-dev.ces-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2f41828262170dab4232395352c9ee3abe25e583
--- /dev/null
+++ b/dev/tatoeba-dev.ces-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1f551413fcedc5a9473fd07828b795e7ede23d88d328b607e6344c9fae5d09e
+size 78295
diff --git a/dev/tatoeba-dev.ces-hun.tsv b/dev/tatoeba-dev.ces-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d12439e8e5d0fac7bfb45119224d5734f9cde85a
--- /dev/null
+++ b/dev/tatoeba-dev.ces-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a2ff1ca086b7b796068c3be268febe6f07d02e9be28f402d5cc80d9e141bb70
+size 6685
diff --git a/dev/tatoeba-dev.ces-rus.tsv b/dev/tatoeba-dev.ces-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f51a282f377396a3bcd0959fd5b5eed29677b849
--- /dev/null
+++ b/dev/tatoeba-dev.ces-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ad518aa7314d958efef12cb24b75ec726b1f86261722edaf4847b411c56d07b
+size 474691
diff --git a/dev/tatoeba-dev.ces-spa.tsv b/dev/tatoeba-dev.ces-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..46eae81951e1b2b8567342e29c3fa8b66f271410
--- /dev/null
+++ b/dev/tatoeba-dev.ces-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbca92db0492124abedbd3d7cde2ae7874836241a31c5f34650a4b3ffe815ace
+size 8887
diff --git a/dev/tatoeba-dev.ces-ukr.tsv b/dev/tatoeba-dev.ces-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..217a5819e89fe95f4c9da2577bfd2a66e64161c8
--- /dev/null
+++ b/dev/tatoeba-dev.ces-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c7bbed81a9a2c1d848331d0c77476faefaee36e378de5b833211581fa105681
+size 81016
diff --git a/dev/tatoeba-dev.chm-rus.tsv b/dev/tatoeba-dev.chm-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..500651a650a874aa15cb79bda92ee164d6edacdd
--- /dev/null
+++ b/dev/tatoeba-dev.chm-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c64db8c30d208b16dcc5c21f0b772103f55d688a02256b9b69d61568229fe6a
+size 133979
diff --git a/dev/tatoeba-dev.cmn_Hans-lzh_Hani.tsv b/dev/tatoeba-dev.cmn_Hans-lzh_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c0ddd45c0607c9124b431a818fdfc612a33a6452
--- /dev/null
+++ b/dev/tatoeba-dev.cmn_Hans-lzh_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:823ca60a3bf05534f008adda7ca9b0df9f959585776e4804467885f300e14509
+size 2778
diff --git a/dev/tatoeba-dev.cmn_Hans-wuu.tsv b/dev/tatoeba-dev.cmn_Hans-wuu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e7614f3f93b5fe96b61cdadb15061a3da74b0cda
--- /dev/null
+++ b/dev/tatoeba-dev.cmn_Hans-wuu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:075b9e2805dcd16fecf064735d278e0d28aa2ce78e2899f905e47ac3d450c572
+size 162314
diff --git a/dev/tatoeba-dev.cmn_Hans-wuu_Hani.tsv b/dev/tatoeba-dev.cmn_Hans-wuu_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bc7211dabd5159c49e7f1b27cd4fbd04dafce25e
--- /dev/null
+++ b/dev/tatoeba-dev.cmn_Hans-wuu_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05eff57d0bbd82249b6b7a93b7e1273cb02d4a1c0e8ccdd8171a7c9b57a672c0
+size 3965
diff --git a/dev/tatoeba-dev.cmn_Hans-yue_Hans.tsv b/dev/tatoeba-dev.cmn_Hans-yue_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..140fc0bc2c696c1edfc0c930ac474755d43d0167
--- /dev/null
+++ b/dev/tatoeba-dev.cmn_Hans-yue_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a25c07f26b9767edc283002cd8d192320f75f4c13caf8c3cf7ced38ebec5094a
+size 10500
diff --git a/dev/tatoeba-dev.cmn_Hans-yue_Hant.tsv b/dev/tatoeba-dev.cmn_Hans-yue_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0febe65029c3fa91eb527e2afc6a3f72a9fd35c5
--- /dev/null
+++ b/dev/tatoeba-dev.cmn_Hans-yue_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60cd2c10cec6bf5809f60224230c731660e67afd951ca37f6d34468c54f34b5d
+size 5759
diff --git a/dev/tatoeba-dev.cmn_Hant-lzh_Hani.tsv b/dev/tatoeba-dev.cmn_Hant-lzh_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..09f74d4a7b1d88d9f66abef93d63683321348012
--- /dev/null
+++ b/dev/tatoeba-dev.cmn_Hant-lzh_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e12d0cdc736fa0d61fc209e2f439611abfd6c754055344b03ed16e3f4440dc8
+size 3392
diff --git a/dev/tatoeba-dev.cmn_Hant-wuu.tsv b/dev/tatoeba-dev.cmn_Hant-wuu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ede7935ece8da1ae41ffd5522b1549f9c271f939
--- /dev/null
+++ b/dev/tatoeba-dev.cmn_Hant-wuu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ffa4a4283f91ddf7c487187932d11bcc7031f219b9884af953e6481edeb9b73
+size 28679
diff --git a/dev/tatoeba-dev.cmn_Hant-wuu_Hani.tsv b/dev/tatoeba-dev.cmn_Hant-wuu_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b7bece097fdc39f9b79ed28857288946b10de506
--- /dev/null
+++ b/dev/tatoeba-dev.cmn_Hant-wuu_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14e2db14a56bf328e2029f38cf3113a67d4330f13c546a4a0984bb89c2f6192b
+size 9505
diff --git a/dev/tatoeba-dev.cmn_Hant-yue_Hans.tsv b/dev/tatoeba-dev.cmn_Hant-yue_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fc34f9f2c1b1b1670db2f25debac186dcce455e9
--- /dev/null
+++ b/dev/tatoeba-dev.cmn_Hant-yue_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6d5aa055bddb11ccf5f51db270167b299f5b41a63edb1d3159a3dc9a47ce3a4
+size 15282
diff --git a/dev/tatoeba-dev.cmn_Hant-yue_Hant.tsv b/dev/tatoeba-dev.cmn_Hant-yue_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4a13baaf7c3a489efd62334c291187179305029e
--- /dev/null
+++ b/dev/tatoeba-dev.cmn_Hant-yue_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf01ed4eae666fddd77a3e875d8f2c247ee747bb954f9e30321913b30c9ffa15
+size 6557
diff --git a/dev/tatoeba-dev.cor-eng.tsv b/dev/tatoeba-dev.cor-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e5bcfa408dfde0399fa7701b116035389dce8f0c
--- /dev/null
+++ b/dev/tatoeba-dev.cor-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:154f711bcbda804da77e14ef822c85aab6f1d3f9bcc95dc53f6b4f2357f0610b
+size 51376
diff --git a/dev/tatoeba-dev.dan-deu.tsv b/dev/tatoeba-dev.dan-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cf5f174cbfc1dec53904586cc000612a76683a1b
--- /dev/null
+++ b/dev/tatoeba-dev.dan-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24bbee940dff12e5c4b762569d9cf80fd434cd4d848fd684cc3c46e2edcfeb61
+size 570378
diff --git a/dev/tatoeba-dev.dan-eng.tsv b/dev/tatoeba-dev.dan-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4451e9c29c5999c308d3119aaf3556e51f1a525b
--- /dev/null
+++ b/dev/tatoeba-dev.dan-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f97d728839470d01ba486436644335dc568ff124511169471a45924d836ccf0
+size 1491198
diff --git a/dev/tatoeba-dev.dan-epo.tsv b/dev/tatoeba-dev.dan-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b83df0c288473d56384e1d61c8852596f3c937ae
--- /dev/null
+++ b/dev/tatoeba-dev.dan-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9baa0bf1819f416d79353c02429bc05a27c750d32f9a293515522ade0c9c9a13
+size 1476168
diff --git a/dev/tatoeba-dev.dan-fin.tsv b/dev/tatoeba-dev.dan-fin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f1ec462a948b124b067cfc29e3340970b36f2e71
--- /dev/null
+++ b/dev/tatoeba-dev.dan-fin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac7b508d72111235aa246bf2fab82e6489ab02a09f11768d2a8b8ae75f7bf0b8
+size 125290
diff --git a/dev/tatoeba-dev.dan-fra.tsv b/dev/tatoeba-dev.dan-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d13713404274c6229909fa3ff899aeafa16e0c4c
--- /dev/null
+++ b/dev/tatoeba-dev.dan-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd57ddd7a2d0125b610fc48baf48407720cbbc81a5c589eb96cb984927303aa5
+size 4617
diff --git a/dev/tatoeba-dev.dan-nld.tsv b/dev/tatoeba-dev.dan-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0fd57dfd1f16f334a2e69b9e2d0f66df9140ad48
--- /dev/null
+++ b/dev/tatoeba-dev.dan-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1827cb5f717f4e5fbefcd678b52b9347c3e9b3fcbc21b00aec7ea7c33a137f7f
+size 2485
diff --git a/dev/tatoeba-dev.dan-nob.tsv b/dev/tatoeba-dev.dan-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..022dece129a374345229b20dd5d05bb956492b96
--- /dev/null
+++ b/dev/tatoeba-dev.dan-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15ab465b82e9e8a140bddfca7582e60eb97ec23ccb28f0c1d1c12c4fa80debe6
+size 79099
diff --git a/dev/tatoeba-dev.dan-nor.tsv b/dev/tatoeba-dev.dan-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bedb970355e040da18f4563d7a977ce2b97a470a
--- /dev/null
+++ b/dev/tatoeba-dev.dan-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c313db80e4294a11ee5cbd1ee37b2b0dc614716cbbbb35e1a1120826ea4310f
+size 79691
diff --git a/dev/tatoeba-dev.dan-rus.tsv b/dev/tatoeba-dev.dan-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8051e7056015612dcbac394dd65626a9b3e707cc
--- /dev/null
+++ b/dev/tatoeba-dev.dan-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:087b34604870819751b983dd7c66d5e8196f6d7e2434058eac8a1b201aed7098
+size 92500
diff --git a/dev/tatoeba-dev.dan-spa.tsv b/dev/tatoeba-dev.dan-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e3a9cbedfe5beb82412aa1be9226f79bff05e5a2
--- /dev/null
+++ b/dev/tatoeba-dev.dan-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:711c4184c6116af170951a4080f34194e309dac987b7cadb65ba9b8a6492dd5c
+size 389803
diff --git a/dev/tatoeba-dev.dan-swe.tsv b/dev/tatoeba-dev.dan-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5b3735533517b336c0d1cafa202c94c44ea3ca5d
--- /dev/null
+++ b/dev/tatoeba-dev.dan-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb5540bc84ceb52ca9e6c8fa20c2d0ac93e79b9b257ac15d77ea4f743b8f55eb
+size 98464
diff --git a/dev/tatoeba-dev.deu-cmn_Hans.tsv b/dev/tatoeba-dev.deu-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..76baf433c55dfd4cb37a03c40fde93cd78739385
--- /dev/null
+++ b/dev/tatoeba-dev.deu-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c9cc7b26742bf64d468ca36c8cba64cc609799b36108f9ea51dea85aa81446c
+size 385605
diff --git a/dev/tatoeba-dev.deu-cmn_Hant.tsv b/dev/tatoeba-dev.deu-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c57045cbede599ed33be78d3cb7d86d402e9403a
--- /dev/null
+++ b/dev/tatoeba-dev.deu-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75f7c4190b0c0372ef0293cdc87121a11006b677b863ccc5bbd1c6c72f163e35
+size 172171
diff --git a/dev/tatoeba-dev.deu-deu.tsv b/dev/tatoeba-dev.deu-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b462c5e5e510f098baa3a90710d0fefd02bde60d
--- /dev/null
+++ b/dev/tatoeba-dev.deu-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cb2af35ff3ac8f48ebe6e76af69753e04072dad6f212c7e70ef804b992a5e24
+size 197902
diff --git a/dev/tatoeba-dev.deu-ell.tsv b/dev/tatoeba-dev.deu-ell.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..28b0e92661b48efc07e164231c27f26e136af200
--- /dev/null
+++ b/dev/tatoeba-dev.deu-ell.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c48f8be93376571f3f6b0598e7ae512e0d75dbedd942ce3ac116ca382ff8226
+size 273320
diff --git a/dev/tatoeba-dev.deu-eng.tsv b/dev/tatoeba-dev.deu-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..aa58b5f974ddffea1bf500e088b67b2dd6b0c60d
--- /dev/null
+++ b/dev/tatoeba-dev.deu-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8b4e7c1eb6d629cc9a9978041c8f2aaca09fb879d2f36bf919bae1b0d271c90
+size 25137584
diff --git a/dev/tatoeba-dev.deu-epo.tsv b/dev/tatoeba-dev.deu-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2c9c945a47b4bf82afad06b832851443f900c764
--- /dev/null
+++ b/dev/tatoeba-dev.deu-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5f732f8002baf38645a3195024dfd0a49f3a559c90afe4634802c386414fa4b
+size 22254500
diff --git a/dev/tatoeba-dev.deu-fas.tsv b/dev/tatoeba-dev.deu-fas.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e136fba41a4a2c2fd473988f7c58d3f2e40fc3c0
--- /dev/null
+++ b/dev/tatoeba-dev.deu-fas.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e44836f3b9bddb488fef2edb823924816401e61584fcf1618e7c2e142c9e9211
+size 117430
diff --git a/dev/tatoeba-dev.deu-fin.tsv b/dev/tatoeba-dev.deu-fin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e394d9ce2c2b5b64347ca013026f3ca97533f835
--- /dev/null
+++ b/dev/tatoeba-dev.deu-fin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee94ab2fd7a1cb5595ed92738668e39925513a7bdba635131c14eb3cc06e63c0
+size 557059
diff --git a/dev/tatoeba-dev.deu-fra.tsv b/dev/tatoeba-dev.deu-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ab038b1435bdb5011678d21f13e6bb4a392fef8b
--- /dev/null
+++ b/dev/tatoeba-dev.deu-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6fe66b6f8580d9e5faf8f2b32be931f1f96a32ffd2457abc6263da6d882e19f7
+size 8942848
diff --git a/dev/tatoeba-dev.deu-heb.tsv b/dev/tatoeba-dev.deu-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0c78ee132f3a9eb2caf7ec5029eb7f9e0fadb448
--- /dev/null
+++ b/dev/tatoeba-dev.deu-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a540c4dfc3436712c0f34c79ce896484502f417bac7dfc60b6bc1e6abc33e9d7
+size 112355
diff --git a/dev/tatoeba-dev.deu-hun.tsv b/dev/tatoeba-dev.deu-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8dfb9f645183c206d2dad8ad89d8b38b2072bb99
--- /dev/null
+++ b/dev/tatoeba-dev.deu-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:298fcfb2bbe31aa5c1f3d5550f18419c83de265ef4af71bb6a0b794c9e0ec4cc
+size 4698144
diff --git a/dev/tatoeba-dev.deu-ido.tsv b/dev/tatoeba-dev.deu-ido.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..89205dd1a0b03433e3b8ab530f76e9b00539442c
--- /dev/null
+++ b/dev/tatoeba-dev.deu-ido.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1598729e71197aea4bd048fe22fadbcaedbccd99bd22aad14ea4ac7b646e368
+size 5979
diff --git a/dev/tatoeba-dev.deu-ile.tsv b/dev/tatoeba-dev.deu-ile.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..751fe3066f2b5faba808f75ff3cfc717ea239bce
--- /dev/null
+++ b/dev/tatoeba-dev.deu-ile.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:535209ac7ff441a7f93ea5dcb101a4df8a1e0be19f92447a91e151a298a2929d
+size 91466
diff --git a/dev/tatoeba-dev.deu-ina.tsv b/dev/tatoeba-dev.deu-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fde41de8dd92ced33090effdfa0d01e269de6d1b
--- /dev/null
+++ b/dev/tatoeba-dev.deu-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f137c9e80d80de579df62100969a73e32a0ebf6beaf3976d4195ec5ba90b0df0
+size 8179
diff --git a/dev/tatoeba-dev.deu-ita.tsv b/dev/tatoeba-dev.deu-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7fa8824c1ae62d2db6b9fe5640fe3ee34cb140e4
--- /dev/null
+++ b/dev/tatoeba-dev.deu-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50ff9f20fc502abed4b166a065c08553f9b354934dd33a45954de0671d579a82
+size 1050058
diff --git a/dev/tatoeba-dev.deu-jpn.tsv b/dev/tatoeba-dev.deu-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ff3930e3704d0c3f852446ce85bb380522b94e3a
--- /dev/null
+++ b/dev/tatoeba-dev.deu-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b690608e215394a9752d08fda58ec1ec7a651c6080fafd0b16405082b18748a7
+size 3232720
diff --git a/dev/tatoeba-dev.deu-jpn_Hani.tsv b/dev/tatoeba-dev.deu-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3a363d41f1964ddc7d37df6f7be23e21ba2f7ce2
--- /dev/null
+++ b/dev/tatoeba-dev.deu-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b91a4f0ce4f3226c894587e7003dfbfb4de66fbb085a02de1c252a22ce925f37
+size 355780
diff --git a/dev/tatoeba-dev.deu-jpn_Hira.tsv b/dev/tatoeba-dev.deu-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7b930ed5dee5d85e158e544eeb3dd25fe0f00aa1
--- /dev/null
+++ b/dev/tatoeba-dev.deu-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ec50a035311d644b73fd9dbf4853acf947771e607811367d254e5bd66960dc6
+size 2787804
diff --git a/dev/tatoeba-dev.deu-jpn_Kana.tsv b/dev/tatoeba-dev.deu-jpn_Kana.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..18f0534feb10bd5d3ce78065b72199dcf9b14a5b
--- /dev/null
+++ b/dev/tatoeba-dev.deu-jpn_Kana.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03e0d96aa3e3652d6ce900ac8fc7b54a13d3f2da6c5bfb5486b7d252f3cf7d84
+size 86656
diff --git a/dev/tatoeba-dev.deu-lad.tsv b/dev/tatoeba-dev.deu-lad.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..537ae78ae5709e6b35638fffcc2ad6ebd1b749b8
--- /dev/null
+++ b/dev/tatoeba-dev.deu-lad.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f237b80fddee6759c5d494957c78e734719c4dc8a10465edd35118fd7d94126e
+size 2010
diff --git a/dev/tatoeba-dev.deu-lat.tsv b/dev/tatoeba-dev.deu-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..98279786bbe7953d5cf616fa6b5484c725870cfd
--- /dev/null
+++ b/dev/tatoeba-dev.deu-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4af11a33f78e711c79a2bc15d5e55ea0e65f256ef8f840b825138f6969467cdd
+size 81264
diff --git a/dev/tatoeba-dev.deu-lfn.tsv b/dev/tatoeba-dev.deu-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..16787d49422e7bf1e30fabb54cdea1f7c7e1868e
--- /dev/null
+++ b/dev/tatoeba-dev.deu-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:956ed3d8f79f8066a67bd33b7e2ddaeaabea1e0a6dbc541a4c5686d8c919d5cd
+size 5279
diff --git a/dev/tatoeba-dev.deu-lfn_Cyrl.tsv b/dev/tatoeba-dev.deu-lfn_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..da003104da65ea66a79766aab76e92ef4a8adc5e
--- /dev/null
+++ b/dev/tatoeba-dev.deu-lfn_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:664b3cdd72e427e63014eb6f2aedaea781f8c28b88fed206e31e6b16cb837598
+size 2652
diff --git a/dev/tatoeba-dev.deu-lfn_Latn.tsv b/dev/tatoeba-dev.deu-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..998498ffe9d80793cdcc208cb908c68a5ff1b808
--- /dev/null
+++ b/dev/tatoeba-dev.deu-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a290273d28553399aff35b9426c8ad4bd724e06e7de2ea1859f2972a2c328c8b
+size 2627
diff --git a/dev/tatoeba-dev.deu-lit.tsv b/dev/tatoeba-dev.deu-lit.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..81be3557092fcab25f771112205531b140c9c319
--- /dev/null
+++ b/dev/tatoeba-dev.deu-lit.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68f1f087e9f0b5e5f81320eeae739b4e81fd32b47a7bc797bcaa480da06126fd
+size 25303
diff --git a/dev/tatoeba-dev.deu-lzh_Hani.tsv b/dev/tatoeba-dev.deu-lzh_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..efd4526d067cad479175bff535799f195e63755d
--- /dev/null
+++ b/dev/tatoeba-dev.deu-lzh_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:140af8787efc47a22a3ea3db91e04186f344781e62b4a4a6f5c1816e07432c37
+size 1780
diff --git a/dev/tatoeba-dev.deu-nds.tsv b/dev/tatoeba-dev.deu-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..44a11a6d96b228f78e31ffd04af90aa89d423370
--- /dev/null
+++ b/dev/tatoeba-dev.deu-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc11a389ace42a1e615d6ee2e7086ec887c8c9325497b721805448b578d15b2c
+size 630733
diff --git a/dev/tatoeba-dev.deu-nld.tsv b/dev/tatoeba-dev.deu-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1dd452152dca78464f68ad07abd7995d55b174ba
--- /dev/null
+++ b/dev/tatoeba-dev.deu-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f31687efae7ec31be2cece277b24b3569dc10e477ea77e1e1525cb8bd4b51f5
+size 2091548
diff --git a/dev/tatoeba-dev.deu-nno.tsv b/dev/tatoeba-dev.deu-nno.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6087b99d4876e43b432dcd94e586d023340cd6c0
--- /dev/null
+++ b/dev/tatoeba-dev.deu-nno.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63755ef3ed274b09147cc94c7ca1e821812d37e4a7a8d1c7cc1de34ba58e0ea1
+size 2662
diff --git a/dev/tatoeba-dev.deu-nob.tsv b/dev/tatoeba-dev.deu-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d5d8c6d885c7626aac525fc9380473f3d04ec4a9
--- /dev/null
+++ b/dev/tatoeba-dev.deu-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7facdb1a7072bcc508153d8091eb67b03d376712fd18b92515ef24bcb0fd935e
+size 97801
diff --git a/dev/tatoeba-dev.deu-nor.tsv b/dev/tatoeba-dev.deu-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..42625953d043acf68c2b901aef75d5574bdb2dae
--- /dev/null
+++ b/dev/tatoeba-dev.deu-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5eb5c05eb124e1b063a24461bfee9d8b639e2595e0b214931e54335cbfc54751
+size 100463
diff --git a/dev/tatoeba-dev.deu-pol.tsv b/dev/tatoeba-dev.deu-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fd6c9ee79c8679b8e5d74750c7390df8488ac5f3
--- /dev/null
+++ b/dev/tatoeba-dev.deu-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab554fab94a47bd5861530e27d21bbb7aad8006de184efb18bbcbbf98ca5f756
+size 461017
diff --git a/dev/tatoeba-dev.deu-por.tsv b/dev/tatoeba-dev.deu-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..93c961900b91334367e988b0bbbe4f2b2af1e60d
--- /dev/null
+++ b/dev/tatoeba-dev.deu-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff25a6eabc5758c5c17395ed4795cdaabaa1691eed6956dfaa11133b95db5ff3
+size 651934
diff --git a/dev/tatoeba-dev.deu-ron.tsv b/dev/tatoeba-dev.deu-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..94f7b4295f5224de3b54d0c70d3297a9386e9edc
--- /dev/null
+++ b/dev/tatoeba-dev.deu-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a32564cfcec094e0d8dd984759660f3ef3857bfedb941a1a53a34dbb84c51b94
+size 74503
diff --git a/dev/tatoeba-dev.deu-run.tsv b/dev/tatoeba-dev.deu-run.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..61255c2004e86018bc5a3ec3555291ffa90159ec
--- /dev/null
+++ b/dev/tatoeba-dev.deu-run.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a129c74491ba9a0d474fbe302cc0bf2e69724fa391327349b61670a878f91d0
+size 74444
diff --git a/dev/tatoeba-dev.deu-rus.tsv b/dev/tatoeba-dev.deu-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3bec0963909ee64437af7e3a90fe20d0ed3d9e96
--- /dev/null
+++ b/dev/tatoeba-dev.deu-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ddceaee57b70311d9bd011982ce9ff7a1252fc7baf7af7eb3582ec40bbd90525
+size 10371615
diff --git a/dev/tatoeba-dev.deu-spa.tsv b/dev/tatoeba-dev.deu-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..575ec993f96df96d17623a189339212b31f736e8
--- /dev/null
+++ b/dev/tatoeba-dev.deu-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1b424cebdf029aa1d487b1ed8ecb990429eecfa2d7d72b68f5de9cf27f05cd3
+size 6578781
diff --git a/dev/tatoeba-dev.deu-swe.tsv b/dev/tatoeba-dev.deu-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..215f22b4a398b4533a206cede18b27bab38dbd60
--- /dev/null
+++ b/dev/tatoeba-dev.deu-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7667626690f288eaa1a2ce829e8b3c0265168cf2c38ef7c0b1d88f10beb9faa1
+size 84067
diff --git a/dev/tatoeba-dev.deu-swg.tsv b/dev/tatoeba-dev.deu-swg.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bd89a4c2456c3b649ca6db49b42dbb3f2ea0ff8e
--- /dev/null
+++ b/dev/tatoeba-dev.deu-swg.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:090c1d56ab6a1143813ede8b942f14eaa5e539f32819d6da56f096154fec768f
+size 16228
diff --git a/dev/tatoeba-dev.deu-tat.tsv b/dev/tatoeba-dev.deu-tat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6e4c07616464032363dfb410b325f74936e4432e
--- /dev/null
+++ b/dev/tatoeba-dev.deu-tat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7895061f92698d02368eef856706ccfe457a60739d837d000681ddd1c41009ec
+size 511413
diff --git a/dev/tatoeba-dev.deu-tat_Latn.tsv b/dev/tatoeba-dev.deu-tat_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5c4bbb70b77e20df27465b3a276cc6549b333dc5
--- /dev/null
+++ b/dev/tatoeba-dev.deu-tat_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0eaa25af38742394d5115b2a091f2c539d868bbffc1dd9a5482d73fdddc512ea
+size 25644
diff --git a/dev/tatoeba-dev.deu-tlh.tsv b/dev/tatoeba-dev.deu-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8d5a34ae6560bc103a1c1b0b0f5cbc0c85ed5636
--- /dev/null
+++ b/dev/tatoeba-dev.deu-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c5ad34ba174158691ccade7893f434090b6555dcc1d832336533b242767cee8
+size 72280
diff --git a/dev/tatoeba-dev.deu-toki.tsv b/dev/tatoeba-dev.deu-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..02341beba87495c4fa34fee27b55dfd537f5ebf1
--- /dev/null
+++ b/dev/tatoeba-dev.deu-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84e547cc903be7c60b9511b3f4e360ef90d41614cfa65f85886757595d10a83a
+size 1278234
diff --git a/dev/tatoeba-dev.deu-tur.tsv b/dev/tatoeba-dev.deu-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f42edfffbffcb1fcd396fd9002719bc58e82437f
--- /dev/null
+++ b/dev/tatoeba-dev.deu-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7019b564414b5cb28d4303d65b8a0bd29219a00bf5c1226b672d35dbd4f6585
+size 593893
diff --git a/dev/tatoeba-dev.deu-ukr.tsv b/dev/tatoeba-dev.deu-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7d82029df79ab6fb5667cddea179511164a66c3b
--- /dev/null
+++ b/dev/tatoeba-dev.deu-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b17187d09b8bd5e1717e4b112231adef31f52f86ca96fdee1634352bdaf594b6
+size 971543
diff --git a/dev/tatoeba-dev.deu-yid.tsv b/dev/tatoeba-dev.deu-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e08d2f049d72b4bc4fcc9cb1cb84cb124d1f1b4e
--- /dev/null
+++ b/dev/tatoeba-dev.deu-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c0566b332cb2d16cadcea615737ac82179c737c88f2a7969cabd4b56f609845
+size 13155
diff --git a/dev/tatoeba-dev.deu-yue_Hans.tsv b/dev/tatoeba-dev.deu-yue_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5197da1693cb4193fb0f982343f970bb630b93d6
--- /dev/null
+++ b/dev/tatoeba-dev.deu-yue_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c811ef8d2fbd643ee5abfc64fb70ad9206ecfe595391946132ff47d1c431659
+size 5847
diff --git a/dev/tatoeba-dev.deu-yue_Hant.tsv b/dev/tatoeba-dev.deu-yue_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..51ccf86950eeaa0678e608de8656b9122728c76f
--- /dev/null
+++ b/dev/tatoeba-dev.deu-yue_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b119d78f27701fef51972521a16ed2d1f0a5f6479804aacc407823f8e066327
+size 6205
diff --git a/dev/tatoeba-dev.deu-zho.tsv b/dev/tatoeba-dev.deu-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4083fd29f35c5e7f4b14f64ef8e342c0f8ba4330
--- /dev/null
+++ b/dev/tatoeba-dev.deu-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b010527d414769f368e852308ffc970ae8f72525d299c3335430bd850474733a
+size 574107
diff --git a/dev/tatoeba-dev.dtp-eng.tsv b/dev/tatoeba-dev.dtp-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5aa03cd8c09b098f3b4cfcd55656412f477defc5
--- /dev/null
+++ b/dev/tatoeba-dev.dtp-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2229c145a0c8318382be491dd64c0a99cb9c93ec82a89a1eaf3bdb9dd3b793b
+size 77040
diff --git a/dev/tatoeba-dev.ell-eng.tsv b/dev/tatoeba-dev.ell-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..512a7b26ecdd74f699e4761c12ec07cd358c6ffb
--- /dev/null
+++ b/dev/tatoeba-dev.ell-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f951972ee8b13e77b1cb1a68e01822e3b0629a07b5c504618430bca18d23855
+size 1097718
diff --git a/dev/tatoeba-dev.ell-nld.tsv b/dev/tatoeba-dev.ell-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8a2668212dae541a5f505c2b2fae92bba8b1bbe0
--- /dev/null
+++ b/dev/tatoeba-dev.ell-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a44aa67cdd012f0b443de83d701e6ff37f15d40843a718fd8c3b5376e765a1a
+size 9164
diff --git a/dev/tatoeba-dev.ell-por.tsv b/dev/tatoeba-dev.ell-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cb06b522202db1b578af280bfff476c34cb72c83
--- /dev/null
+++ b/dev/tatoeba-dev.ell-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9bc2299279228a3d6f03563a1522d4d8982eeab55cd300a0910b75dc0ced912f
+size 1845
diff --git a/dev/tatoeba-dev.ell-rus.tsv b/dev/tatoeba-dev.ell-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..872c0a75ef76cdd61557f2599eb3cce106096bec
--- /dev/null
+++ b/dev/tatoeba-dev.ell-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0c9c049d2b321a5743d9b2989ecdac270d842c8f6d7b2ddd3d879a569d033e8
+size 284857
diff --git a/dev/tatoeba-dev.ell-spa.tsv b/dev/tatoeba-dev.ell-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b8f60298a7752b3a813f5a34730b4ffb8d0d1e0e
--- /dev/null
+++ b/dev/tatoeba-dev.ell-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f06d21f102f481c471d7c4461b54911fdc0cc55675aaf5b9bf777013c2c4658
+size 85426
diff --git a/dev/tatoeba-dev.ell-tur.tsv b/dev/tatoeba-dev.ell-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..223bfa7af897aabdedf53ac314c8dab3d077eb9a
--- /dev/null
+++ b/dev/tatoeba-dev.ell-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4047966498414773d0ad466e89e96dc6257aacccf38f2ba2ca83242579598a9a
+size 9099
diff --git a/dev/tatoeba-dev.eng-bos_Latn.tsv b/dev/tatoeba-dev.eng-bos_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..28cbb3545d25f4ad5b185834ab26b01dd2b66094
--- /dev/null
+++ b/dev/tatoeba-dev.eng-bos_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28800ac64433450ab30433d6b10fdeffc9cf125d87ca20e33e01154770b4a03e
+size 12930
diff --git a/dev/tatoeba-dev.eng-ckb.tsv b/dev/tatoeba-dev.eng-ckb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7462b96f656b6a57b0110d5e7855ee043b895e8a
--- /dev/null
+++ b/dev/tatoeba-dev.eng-ckb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5476864f6e558f2b3b9d8f436a81d33634fa6671d6b6676f70f3f3bbac31ad8
+size 1952
diff --git a/dev/tatoeba-dev.eng-cmn_Hans.tsv b/dev/tatoeba-dev.eng-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c2e616d4eeafebf97df2ab58955c37ced01b8436
--- /dev/null
+++ b/dev/tatoeba-dev.eng-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f476d765f2e1db41be8cc617b9e4b669d45ee766b4bcaf30102ac39e8b2ac43e
+size 1561687
diff --git a/dev/tatoeba-dev.eng-cmn_Hant.tsv b/dev/tatoeba-dev.eng-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bc4bf3d2405cd0edad0d70be04b37ecceb3cb7ac
--- /dev/null
+++ b/dev/tatoeba-dev.eng-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48ce7ba1087dd2852f1c28af939c5cd5863377779c4c03fc47ff93235ac1356d
+size 1579337
diff --git a/dev/tatoeba-dev.eng-cmn_Latn.tsv b/dev/tatoeba-dev.eng-cmn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f04b3009c30a65b48e539473b4b537f499c9aa92
--- /dev/null
+++ b/dev/tatoeba-dev.eng-cmn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7426c2027058f502a515ad76b3f778c4584c5c11714feb18040dcc68f6e9ec1
+size 2597
diff --git a/dev/tatoeba-dev.eng-eng.tsv b/dev/tatoeba-dev.eng-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d5d4a1f7a5498f7f143e4f8cc36505a73b86b92a
--- /dev/null
+++ b/dev/tatoeba-dev.eng-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:637a37db978daa88cfbdf706e16b540229bf0ef98064c1a78f62288904325e23
+size 9036672
diff --git a/dev/tatoeba-dev.eng-epo.tsv b/dev/tatoeba-dev.eng-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..03b2622ce3973de4cdf01ee0dcec69caa213215b
--- /dev/null
+++ b/dev/tatoeba-dev.eng-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3343790c802f63513e6c6bd6a145edd89b8721a37dc9b30e9e0b7142cb319410
+size 19513865
diff --git a/dev/tatoeba-dev.eng-est.tsv b/dev/tatoeba-dev.eng-est.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..924f92f7babd7f72033ac87d6d6e81d648e8bc64
--- /dev/null
+++ b/dev/tatoeba-dev.eng-est.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2142348f129cd6326d8d95b0f2afe168a2f4c0037662099b7a7679412c7ae32d
+size 72275
diff --git a/dev/tatoeba-dev.eng-eus.tsv b/dev/tatoeba-dev.eng-eus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d7a6a081af4a7e1e54550cfc4fd518ee968ba3d8
--- /dev/null
+++ b/dev/tatoeba-dev.eng-eus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3eb08ab723fcd1dffa6610d9fb7a7b344d3846660a25f74bfa73278d575983da
+size 79946
diff --git a/dev/tatoeba-dev.eng-fas.tsv b/dev/tatoeba-dev.eng-fas.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8c772cffccaf25decb5d56aabe07644682f7212f
--- /dev/null
+++ b/dev/tatoeba-dev.eng-fas.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cfa55c03e7916d1c1f3512d88f80c10934c2ff215bf5c173b2a9454d67853d8
+size 112144
diff --git a/dev/tatoeba-dev.eng-fin.tsv b/dev/tatoeba-dev.eng-fin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5234276c1e1e6424bcce00248ba35ae867279f8d
--- /dev/null
+++ b/dev/tatoeba-dev.eng-fin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1a70954fd10303d3220d7f6015bc3797a0b0a979bcb80a7f341295d836b473f
+size 5427786
diff --git a/dev/tatoeba-dev.eng-fra.tsv b/dev/tatoeba-dev.eng-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8cbec9640fce35c66c7ff816ccd954246ad667d8
--- /dev/null
+++ b/dev/tatoeba-dev.eng-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61e1f58c390584647f7a635e51bb4c5bebe1534e7044c9999d6d322b308f31da
+size 21044945
diff --git a/dev/tatoeba-dev.eng-gle.tsv b/dev/tatoeba-dev.eng-gle.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4acf06a1c8f97398600e2479a62e8b53b0f960a1
--- /dev/null
+++ b/dev/tatoeba-dev.eng-gle.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c941a9f0d9f1f7263b0219af37ad5d6327db6d349eb610095b7dcbcd1e310ab
+size 2051
diff --git a/dev/tatoeba-dev.eng-gos.tsv b/dev/tatoeba-dev.eng-gos.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cd77fe7e4052ac5f9570a9d5e492ffe6f3530e80
--- /dev/null
+++ b/dev/tatoeba-dev.eng-gos.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:710f2d236f706ca65e6170769c1a5ae4f16cdc2fd9fa3d4f2d6e48511d70e9c7
+size 6300
diff --git a/dev/tatoeba-dev.eng-hbs.tsv b/dev/tatoeba-dev.eng-hbs.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3674da36f716d1cdc01fd73889c79173b9e8587e
--- /dev/null
+++ b/dev/tatoeba-dev.eng-hbs.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af28931711e84e4c7033ead1c8f377f93d4ae00dd11d0244f6a0d839c6f98d11
+size 1267994
diff --git a/dev/tatoeba-dev.eng-heb.tsv b/dev/tatoeba-dev.eng-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ddd8356f9f5ac2a7f30d9cadae0b535098173315
--- /dev/null
+++ b/dev/tatoeba-dev.eng-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ececad35435b183d4b2ecd1464ea14388ecdf4e08a478428e22d6ddf9898db7
+size 13176238
diff --git a/dev/tatoeba-dev.eng-hin.tsv b/dev/tatoeba-dev.eng-hin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c7c29c97a0ca25fe3999fb15882300a1be98b06f
--- /dev/null
+++ b/dev/tatoeba-dev.eng-hin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0595b5b32d49e207813b3d22ffd60a1804a762be8b9e3c826068c8c622f29341
+size 679432
diff --git a/dev/tatoeba-dev.eng-hrv.tsv b/dev/tatoeba-dev.eng-hrv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f9e692f865ec2e8d0801629847c85a5a7e3294b8
--- /dev/null
+++ b/dev/tatoeba-dev.eng-hrv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52b9fda42048d4ae11d66380a07972c450fef5d3be2f618d55500b8aa77b9062
+size 67797
diff --git a/dev/tatoeba-dev.eng-hun.tsv b/dev/tatoeba-dev.eng-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..81789fed737e111f4ba75a1d7382a41f477c34ba
--- /dev/null
+++ b/dev/tatoeba-dev.eng-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57fa2744c49f9367b3569f864924a2a514be9953ee02da66a31d0bb2bedcda01
+size 6949776
diff --git a/dev/tatoeba-dev.eng-hye.tsv b/dev/tatoeba-dev.eng-hye.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3c76fff51e5d42c54261dbef9f1bba5c355d9bf1
--- /dev/null
+++ b/dev/tatoeba-dev.eng-hye.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d009151bd556828e67b12562b3236dd5ca8016efedb1965b5f2ea50d95af3fa5
+size 66759
diff --git a/dev/tatoeba-dev.eng-ido.tsv b/dev/tatoeba-dev.eng-ido.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..847647e5becb8a972d197329c4b4f150319f673a
--- /dev/null
+++ b/dev/tatoeba-dev.eng-ido.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6afdc970da1bafc884ca01b98153ec6fb4151b6ebe59581da8c3c175944bc6c4
+size 109265
diff --git a/dev/tatoeba-dev.eng-ile.tsv b/dev/tatoeba-dev.eng-ile.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..32080eaf8e0c9d942870ca745c928b6d867ceaee
--- /dev/null
+++ b/dev/tatoeba-dev.eng-ile.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1e340740243f60df99733546179398335b5dd2624a15573021485a577499b99
+size 72927
diff --git a/dev/tatoeba-dev.eng-ilo.tsv b/dev/tatoeba-dev.eng-ilo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1ab006752fe6d6e6df864600316ada299117b4cd
--- /dev/null
+++ b/dev/tatoeba-dev.eng-ilo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85f744b0c957050d34ed5ff504dff5b83be5581daf5baa5a7af87373e873c200
+size 71696
diff --git a/dev/tatoeba-dev.eng-ina.tsv b/dev/tatoeba-dev.eng-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c04f24b277e03f33b0a88341b150aee2666fdfd2
--- /dev/null
+++ b/dev/tatoeba-dev.eng-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc97d5c0bf0c8fd1ef2744321a34c89bed0bc8b4038f225f31e2b2200b3fa427
+size 574881
diff --git a/dev/tatoeba-dev.eng-ind.tsv b/dev/tatoeba-dev.eng-ind.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a42a161e4bb67d2087a328ff5e5fc4cbefe4e0c6
--- /dev/null
+++ b/dev/tatoeba-dev.eng-ind.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b47ba3de89a771135e69043f39b0e4cccdcaf0fec4a247cef5ab7fdf1b2eb1f9
+size 450871
diff --git a/dev/tatoeba-dev.eng-isl.tsv b/dev/tatoeba-dev.eng-isl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d528ec22581c8be22f530ea70b7cd311240d05b7
--- /dev/null
+++ b/dev/tatoeba-dev.eng-isl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a2d384a9780381afa4d49d7eafae8feeadb94b68ace76907bc24ebfaaf6fb6a
+size 579029
diff --git a/dev/tatoeba-dev.eng-ita.tsv b/dev/tatoeba-dev.eng-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a4a0222603b7de554cdd8f55bdee7b82970e07c4
--- /dev/null
+++ b/dev/tatoeba-dev.eng-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:699240062281aa8a768d66772d7e8f9f42b37341e17b5bbbe1deea8f108895f3
+size 32737430
diff --git a/dev/tatoeba-dev.eng-jbo.tsv b/dev/tatoeba-dev.eng-jbo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1671e544db7dad1adb967f79a296e669bc228be3
--- /dev/null
+++ b/dev/tatoeba-dev.eng-jbo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a360662184fe92eda8cef1e4f29175965ba1e6aa202d479c422abca03b64e72
+size 510193
diff --git a/dev/tatoeba-dev.eng-jbo_Latn.tsv b/dev/tatoeba-dev.eng-jbo_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6d21f82634a67d072d7698492e7be4bc2938d589
--- /dev/null
+++ b/dev/tatoeba-dev.eng-jbo_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d56a4dee2135e19c1483456fe3c7505c051311d37dcee29f8cc9d6ddd972cbf
+size 510113
diff --git a/dev/tatoeba-dev.eng-jpn.tsv b/dev/tatoeba-dev.eng-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f486a40b53902578427f7a563696ff84880dda3e
--- /dev/null
+++ b/dev/tatoeba-dev.eng-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4de62963d8cb33ba26cf0294a740331e51a5c84bc79c8e46f2d26219860be8e5
+size 21017389
diff --git a/dev/tatoeba-dev.eng-jpn_Hani.tsv b/dev/tatoeba-dev.eng-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7ef1c648b02bb88471af5cba6c9ffba3efd07e21
--- /dev/null
+++ b/dev/tatoeba-dev.eng-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adf6b6a126c067dcf57692c1a19aaf9abd66cb727d061a6d18f6aaaf7c7914f1
+size 1982821
diff --git a/dev/tatoeba-dev.eng-jpn_Hira.tsv b/dev/tatoeba-dev.eng-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0d06e7c98a2dcdd4c4f04b13be7fa95d5f4e1433
--- /dev/null
+++ b/dev/tatoeba-dev.eng-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcd308456dc3343fa6f351391fb01783ca9797051fbbd302f18126a40df38514
+size 18681476
diff --git a/dev/tatoeba-dev.eng-jpn_Kana.tsv b/dev/tatoeba-dev.eng-jpn_Kana.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c2c0e0004d1a6ce671f55890c7da07e4e897e161
--- /dev/null
+++ b/dev/tatoeba-dev.eng-jpn_Kana.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b672331fbbbfdf7cb69997a4e1713ea80aa3a87e1a33e0ac8178a0bc55012f4d
+size 344882
diff --git a/dev/tatoeba-dev.eng-jpn_Latn.tsv b/dev/tatoeba-dev.eng-jpn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f06012407c74f8068ea365dcec854f02f3a9d4c5
--- /dev/null
+++ b/dev/tatoeba-dev.eng-jpn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0948ba34465459483b5a49fbdc7daaa5b6187e6b08f56a287d3acb3cdfbf126a
+size 6603
diff --git a/dev/tatoeba-dev.eng-kab.tsv b/dev/tatoeba-dev.eng-kab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..db3c4fe8070347949cfc0bd266c5d4b3d7e4dc41
--- /dev/null
+++ b/dev/tatoeba-dev.eng-kab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7d25e19101eccdf95440bb7cb8a23ac740b1020a7dda8794c210e58625a309a
+size 926831
diff --git a/dev/tatoeba-dev.eng-kha.tsv b/dev/tatoeba-dev.eng-kha.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6013dceaa44cde281d24415009ffcfa3e8d0eabf
--- /dev/null
+++ b/dev/tatoeba-dev.eng-kha.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6932d1e172343b1409f002d46c201aeb2ef5dd4e06a0d040e5a687d0c346c05b
+size 1635
diff --git a/dev/tatoeba-dev.eng-kmr_Latn.tsv b/dev/tatoeba-dev.eng-kmr_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0008432731901ffbd3c31f4db5dc74950457a86e
--- /dev/null
+++ b/dev/tatoeba-dev.eng-kmr_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:929522925fb821a9097e3ae2f2bda147e142eac05273c233236c04eb554e807b
+size 3554
diff --git a/dev/tatoeba-dev.eng-kor.tsv b/dev/tatoeba-dev.eng-kor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1c166f6f196c5cd66e48709a8d42f21bfbecbf11
--- /dev/null
+++ b/dev/tatoeba-dev.eng-kor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c115977b9288ed2b834e0500133b9cbddbf03e682bef85d15502805fd587da0b
+size 92025
diff --git a/dev/tatoeba-dev.eng-kor_Hang.tsv b/dev/tatoeba-dev.eng-kor_Hang.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b3473b5f6a6ae36379dd6d442fd1c5ad3b87f85d
--- /dev/null
+++ b/dev/tatoeba-dev.eng-kor_Hang.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44c66023ab7d0f3756b66fefbf8df93573186e1a4bbe1d6d94146c50a3797102
+size 91437
diff --git a/dev/tatoeba-dev.eng-kur.tsv b/dev/tatoeba-dev.eng-kur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..09041365c3e7c0876fcb69d808ab9a1db7d1ee91
--- /dev/null
+++ b/dev/tatoeba-dev.eng-kur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f861f477edf67a71fd8b47c40108f0297836e99d7e0b3ffd33670d784c11e680
+size 5812
diff --git a/dev/tatoeba-dev.eng-lad.tsv b/dev/tatoeba-dev.eng-lad.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a66c7af167a7707c1938e5caf84806f67f0735db
--- /dev/null
+++ b/dev/tatoeba-dev.eng-lad.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4978a6a916104bf18291505661d2f0f6f6742e38ba8f4c7b67866d0c699db56e
+size 3811
diff --git a/dev/tatoeba-dev.eng-lad_Latn.tsv b/dev/tatoeba-dev.eng-lad_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6fab26f6719c81faff4b2e9389520ccf424fe3bf
--- /dev/null
+++ b/dev/tatoeba-dev.eng-lad_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23a49537368eaa18498a795ec2fa6385943a086ebd3a589dca69e603681ff61f
+size 2197
diff --git a/dev/tatoeba-dev.eng-lat.tsv b/dev/tatoeba-dev.eng-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..12d5c26c34cc322daee08e4cf8669a3182f8f8ba
--- /dev/null
+++ b/dev/tatoeba-dev.eng-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1027212834658bfc2ff07812f7ca167ddba7b29e7adbefb4009d52ae7aa9f9d3
+size 1325827
diff --git a/dev/tatoeba-dev.eng-lfn.tsv b/dev/tatoeba-dev.eng-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..539348adca26a3991a6da41618a4bd409afd61ef
--- /dev/null
+++ b/dev/tatoeba-dev.eng-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18fcaa9d23e09f0790bb3fd8a1670a39f2eec3977f022d04b0041bbc9d08b9c6
+size 334527
diff --git a/dev/tatoeba-dev.eng-lfn_Cyrl.tsv b/dev/tatoeba-dev.eng-lfn_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a3ed70b133f67c46b54d63c43ced7c9135aa779d
--- /dev/null
+++ b/dev/tatoeba-dev.eng-lfn_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07811e9baf92d8c123901502c2dddaa67390de9a0d96d3b008aafee9e1fc5441
+size 136461
diff --git a/dev/tatoeba-dev.eng-lfn_Latn.tsv b/dev/tatoeba-dev.eng-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1ff6fba6cf9e92cc98776bdb4068b1e5d97a758c
--- /dev/null
+++ b/dev/tatoeba-dev.eng-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26c174452c7bdf6d2d458cf2cde33eaacab743cc26f4002c3097f67bc69271fc
+size 198066
diff --git a/dev/tatoeba-dev.eng-lit.tsv b/dev/tatoeba-dev.eng-lit.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4b4510ac98548c65f65f54334987691dab35173b
--- /dev/null
+++ b/dev/tatoeba-dev.eng-lit.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a766611d13e5e8c39437dd41f97c87d4f8224c788554d6947b3f53c250e42faf
+size 400501
diff --git a/dev/tatoeba-dev.eng-lzh_Hani.tsv b/dev/tatoeba-dev.eng-lzh_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3046c5e2360545b64dbfe7138e8eb572bd9d864d
--- /dev/null
+++ b/dev/tatoeba-dev.eng-lzh_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5068c8653a56357422fdebc180eaad7a37c1c19b5ecd3a967f1c1145356c5be4
+size 35639
diff --git a/dev/tatoeba-dev.eng-lzh_Hans.tsv b/dev/tatoeba-dev.eng-lzh_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3d8e62a5429de7f2c271d05c64297f4456e55330
--- /dev/null
+++ b/dev/tatoeba-dev.eng-lzh_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38ecf56f350fc623e7fbfb088bd12027e9f8f3e1b58e842c01f476f9a4c8c91d
+size 1343
diff --git a/dev/tatoeba-dev.eng-mar.tsv b/dev/tatoeba-dev.eng-mar.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b9bc42d02442b9fde9e8b90406bd72b549f0e89f
--- /dev/null
+++ b/dev/tatoeba-dev.eng-mar.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57ebfcf6f759340f6cc80bb7e98ac88fae83d0ea0e397c392611545f506fe2cf
+size 4688040
diff --git a/dev/tatoeba-dev.eng-max_Latn.tsv b/dev/tatoeba-dev.eng-max_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0679a9fe3e67890215023441fb50854a333da144
--- /dev/null
+++ b/dev/tatoeba-dev.eng-max_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cd8a0dcf9495b39625dfae6cc41b5a9794df502a07554ff5a9d1890df684071
+size 11067
diff --git a/dev/tatoeba-dev.eng-mkd.tsv b/dev/tatoeba-dev.eng-mkd.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b8a07ded83b4afc51e863b207ec217214962b6b3
--- /dev/null
+++ b/dev/tatoeba-dev.eng-mkd.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63b670415df8fb79df44d611a06d9e9d4ba2aed7e0798b8d6585ff8b90c25077
+size 6073534
diff --git a/dev/tatoeba-dev.eng-mon.tsv b/dev/tatoeba-dev.eng-mon.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c69f2647983a815a5f863f7093039ef3f18651b8
--- /dev/null
+++ b/dev/tatoeba-dev.eng-mon.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4981da4c966a6bbb84b6bd6fe9a1d6e6701d884b5f13c56be50f8ae4caa5891e
+size 2922
diff --git a/dev/tatoeba-dev.eng-msa.tsv b/dev/tatoeba-dev.eng-msa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7937e5f95b932088bc134b62b636e805edc504cd
--- /dev/null
+++ b/dev/tatoeba-dev.eng-msa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f95367b06ab91e3bc7794f997af8fd1c067e3060d4dbdf4dc80c948d651b9597
+size 550682
diff --git a/dev/tatoeba-dev.eng-nds.tsv b/dev/tatoeba-dev.eng-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0a7ec19cccb0949049c00c765a75f51c562c1464
--- /dev/null
+++ b/dev/tatoeba-dev.eng-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7202c9978ba8abd55b9359b105642556032977515728ef645aef47e99bf1d4a2
+size 232162
diff --git a/dev/tatoeba-dev.eng-nld.tsv b/dev/tatoeba-dev.eng-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..429945628d8326e838bd71d6a43f6d47ea377d60
--- /dev/null
+++ b/dev/tatoeba-dev.eng-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c7dfbf38ae844b647a0c84a56170ce4363a8ef9ac865f1de15d2a9a4853bc88
+size 4545762
diff --git a/dev/tatoeba-dev.eng-nno.tsv b/dev/tatoeba-dev.eng-nno.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d5937a233159b21db76cb79f9250a467dd9121eb
--- /dev/null
+++ b/dev/tatoeba-dev.eng-nno.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1506121e427c77823919c68f04ba9ad9209645bbf0cc15ac11cc601e4790faa
+size 38135
diff --git a/dev/tatoeba-dev.eng-nob.tsv b/dev/tatoeba-dev.eng-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..787d13339c01bbc5064a52655917797cea597de4
--- /dev/null
+++ b/dev/tatoeba-dev.eng-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b0436ec7ccb3fb1cfd8ceeb06971b090e0eb17666215d4e7f496a302d278970
+size 420251
diff --git a/dev/tatoeba-dev.eng-nor.tsv b/dev/tatoeba-dev.eng-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..70cfb0e0fd086f1838edbf3c119069ac1dbbb727
--- /dev/null
+++ b/dev/tatoeba-dev.eng-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dea76fe0af7ddc61aa108788b0d1f0cbcc6b261a94e03ecea1707f41ec344a06
+size 458386
diff --git a/dev/tatoeba-dev.eng-pam.tsv b/dev/tatoeba-dev.eng-pam.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1e60b376af4bfe5417bcdf8987abfe1922e33066
--- /dev/null
+++ b/dev/tatoeba-dev.eng-pam.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3665183c2d5b396126503f4e89c19f48f2f021522cbd3413706f33babcd49ba
+size 37820
diff --git a/dev/tatoeba-dev.eng-pol.tsv b/dev/tatoeba-dev.eng-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..02a5fc44f20bbeb94a702523900a3dc761844922
--- /dev/null
+++ b/dev/tatoeba-dev.eng-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b66e39e95aeac3e81e76ffc6551492271ee0847e50a65462a3040b2c4963fe3b
+size 3412012
diff --git a/dev/tatoeba-dev.eng-por.tsv b/dev/tatoeba-dev.eng-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..73cd36a52b3127f75b3bd3b8df605abb520759a4
--- /dev/null
+++ b/dev/tatoeba-dev.eng-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:807fe6d37e3e041282242a0ff0885e395b5eb19074a31f5a3ad9c9506b8c2866
+size 15767111
diff --git a/dev/tatoeba-dev.eng-ron.tsv b/dev/tatoeba-dev.eng-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d594ac9f017fbfba3d4cf9b47c9b13f50846eb0a
--- /dev/null
+++ b/dev/tatoeba-dev.eng-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c55791410027b4d8eac0e788de79c66e150db1e831c52764dcc68f8187b5fafd
+size 778200
diff --git a/dev/tatoeba-dev.eng-rus.tsv b/dev/tatoeba-dev.eng-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6ef3e76a7f85709c1734e0690d3f68dfd4410a50
--- /dev/null
+++ b/dev/tatoeba-dev.eng-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86913d92fe601dfa015e183d4c8f2c6276c062c7ecbecf2cc2dd7668e8097a79
+size 48671339
diff --git a/dev/tatoeba-dev.eng-slv.tsv b/dev/tatoeba-dev.eng-slv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5e53cbc68524b8ba0e7299b473732c29a3fa3028
--- /dev/null
+++ b/dev/tatoeba-dev.eng-slv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf232d01c3d33262e25c4cc12c60e7680254e82db44c56ebd0fac29c87119a18
+size 109906
diff --git a/dev/tatoeba-dev.eng-spa.tsv b/dev/tatoeba-dev.eng-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0c39ac35f725208aa3c58cc8326e2e43d85dac7f
--- /dev/null
+++ b/dev/tatoeba-dev.eng-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f84151a7a4ecc8f8378bedf1bdd6d23302556ab098ecb4e009ffff25e94fd5b0
+size 16211897
diff --git a/dev/tatoeba-dev.eng-srp_Cyrl.tsv b/dev/tatoeba-dev.eng-srp_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..79f7952c65dd9f4d1256bf9c535b767426988262
--- /dev/null
+++ b/dev/tatoeba-dev.eng-srp_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6ea380ac5183d8623ed081f43f214e195b2333480a3db5dc199fb8cfca2d9e1
+size 875779
diff --git a/dev/tatoeba-dev.eng-srp_Latn.tsv b/dev/tatoeba-dev.eng-srp_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..aad5545c8506ddb11da4a4533894baefdf09a3cf
--- /dev/null
+++ b/dev/tatoeba-dev.eng-srp_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c642a574cbce47352ddc7659849116b98faf2184c8ec54160f5ac4e5ca527afd
+size 311488
diff --git a/dev/tatoeba-dev.eng-swe.tsv b/dev/tatoeba-dev.eng-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..acf6dd86383bda28c41e36746e0bff8f49742ec5
--- /dev/null
+++ b/dev/tatoeba-dev.eng-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4d094110e9fd0ab4e02baef165efc83cb3034edb583ca326ea500cd06600a65
+size 1044301
diff --git a/dev/tatoeba-dev.eng-tgl.tsv b/dev/tatoeba-dev.eng-tgl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3a30ab45acb1895fd7004a22907bcae92ed9e0b5
--- /dev/null
+++ b/dev/tatoeba-dev.eng-tgl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44696089d99bd19e8e5b640f8d59dbe09ccb07a913ea783afcfd6aa135861666
+size 387791
diff --git a/dev/tatoeba-dev.eng-tgl_Latn.tsv b/dev/tatoeba-dev.eng-tgl_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b05cda77da61dc127db9f91d29d52512525d94f3
--- /dev/null
+++ b/dev/tatoeba-dev.eng-tgl_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5e4df1282a77295c6efd0aa9fdefba17a276cfb731be42286703f4e1864c116
+size 387736
diff --git a/dev/tatoeba-dev.eng-tlh.tsv b/dev/tatoeba-dev.eng-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6f62a0c0c2587bf70893897215880775b6988523
--- /dev/null
+++ b/dev/tatoeba-dev.eng-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:088e10fe3606d18262e7692d8769e3fcf31301e275540cdcb3f6c6b1c8f127c8
+size 540841
diff --git a/dev/tatoeba-dev.eng-toki.tsv b/dev/tatoeba-dev.eng-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..886cb06a3331079424a0bd78eb15618eef7ec83d
--- /dev/null
+++ b/dev/tatoeba-dev.eng-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:17c311ffa87d4263411bab6f5b52e54251662a4b75c3df52f7c93782a09e2802
+size 709290
diff --git a/dev/tatoeba-dev.eng-tuk.tsv b/dev/tatoeba-dev.eng-tuk.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f6e57b7cff5129c8d9ea47c1a85696897f67bb41
--- /dev/null
+++ b/dev/tatoeba-dev.eng-tuk.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b95515169bed034e3df4b49a69e5c20b4775b9dba3d793298bcab78899f164d1
+size 333180
diff --git a/dev/tatoeba-dev.eng-tur.tsv b/dev/tatoeba-dev.eng-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..29fa0de18fdfc621c2d889c90807e8a11fb94d1b
--- /dev/null
+++ b/dev/tatoeba-dev.eng-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d0e61d3b045a81b2af28b08e9a17ec4c238bf17a68ee58f5619b78495e21297
+size 52917848
diff --git a/dev/tatoeba-dev.eng-tzl.tsv b/dev/tatoeba-dev.eng-tzl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..133614fe6ba258c1a67ec1dcb678476dc1900b91
--- /dev/null
+++ b/dev/tatoeba-dev.eng-tzl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b69dc334ef66ecab34c3374a0b82c791cb93d439d0f974ef19519b516150b626
+size 1313
diff --git a/dev/tatoeba-dev.eng-uig.tsv b/dev/tatoeba-dev.eng-uig.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f87560d9377d3548c189d8853a35e3145bf516cd
--- /dev/null
+++ b/dev/tatoeba-dev.eng-uig.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a0c7c65e136fd2b6d711658f2841e96cd8c8f52e668c227db8fec5054f3cfe5
+size 105302
diff --git a/dev/tatoeba-dev.eng-uig_Arab.tsv b/dev/tatoeba-dev.eng-uig_Arab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f7a573e0734faa05b50918458eaccb797f357c53
--- /dev/null
+++ b/dev/tatoeba-dev.eng-uig_Arab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8b32b17546566440f6ac707ef505ff668de8e9b37c87309878dde6c4451d5e7
+size 105238
diff --git a/dev/tatoeba-dev.eng-ukr.tsv b/dev/tatoeba-dev.eng-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..79a026f19c025781028a136284146a7842e720f4
--- /dev/null
+++ b/dev/tatoeba-dev.eng-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3db507879c617c31dc574aa94f022656b2652b3c884a7fe575fbab5299fa5e8e
+size 13818313
diff --git a/dev/tatoeba-dev.eng-vie.tsv b/dev/tatoeba-dev.eng-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ece0a3550a33ba40108b59b24a393fc3ac2cecf9
--- /dev/null
+++ b/dev/tatoeba-dev.eng-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:060ce02c8192acf3b4f5470e34601dfd34a86f009d669ad87a59b9f27d2fab72
+size 298384
diff --git a/dev/tatoeba-dev.eng-vol.tsv b/dev/tatoeba-dev.eng-vol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bd08a70e10b369a69ad147824f2c2c39f49a6bb6
--- /dev/null
+++ b/dev/tatoeba-dev.eng-vol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10913f77bcef802a8c1b950399214eae442f1f64473a5860efe79167f56ee49d
+size 81012
diff --git a/dev/tatoeba-dev.eng-wuu.tsv b/dev/tatoeba-dev.eng-wuu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..377c895219355ab808c2facfad735befd4a2152b
--- /dev/null
+++ b/dev/tatoeba-dev.eng-wuu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0ea734e2692bd65b25be1c2f87bbf9c9257e8883168c685964ada213c697db6
+size 62255
diff --git a/dev/tatoeba-dev.eng-wuu_Hani.tsv b/dev/tatoeba-dev.eng-wuu_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..31faf2b71db4fbef2232f38e3378bd5d626fe987
--- /dev/null
+++ b/dev/tatoeba-dev.eng-wuu_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e256fa7790c2b8e95628a587978403b276467505cdd93f44b73e2c34f057221f
+size 4158
diff --git a/dev/tatoeba-dev.eng-yid.tsv b/dev/tatoeba-dev.eng-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6b4393e108372ad6944bf1d2f1e8d75a941eba58
--- /dev/null
+++ b/dev/tatoeba-dev.eng-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5b6b4c237d2bc8c70ce8f49ebb1fbd17875dcc4034ec8f48790640b58b74120
+size 180143
diff --git a/dev/tatoeba-dev.eng-yue_Hans.tsv b/dev/tatoeba-dev.eng-yue_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..982a44c3508dac04d74d307abb7f460a448d6df7
--- /dev/null
+++ b/dev/tatoeba-dev.eng-yue_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b0881f84690378617b0fbcf9582695b8bfab38dd3842be6dfbdb4336780968d
+size 258827
diff --git a/dev/tatoeba-dev.eng-yue_Hant.tsv b/dev/tatoeba-dev.eng-yue_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..924387f6b50b8e35e23a29180020246d919eb044
--- /dev/null
+++ b/dev/tatoeba-dev.eng-yue_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:263da23b59b82cd1385adc6e5605866d2e905d51cca6c251a12a51cc2b899e35
+size 113290
diff --git a/dev/tatoeba-dev.eng-zho.tsv b/dev/tatoeba-dev.eng-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..132b093f83a0d874eadf365ee1c357068290c0a0
--- /dev/null
+++ b/dev/tatoeba-dev.eng-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b82735eb6f2349de149f61c96ce4762d031dff76bce409ccd58cc7dcbefb918
+size 3622494
diff --git a/dev/tatoeba-dev.eng-zlm_Latn.tsv b/dev/tatoeba-dev.eng-zlm_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..41c75c6e59ce456d59d6079ab03fb2f684bb72c0
--- /dev/null
+++ b/dev/tatoeba-dev.eng-zlm_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b39de0b3aa224c79dd8bb7df868d86dece8b4040866863fee0c5f2054e6a5922
+size 4247
diff --git a/dev/tatoeba-dev.eng-zsm_Latn.tsv b/dev/tatoeba-dev.eng-zsm_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bdb1bce2cccf9bcbc15f72f580af7333c82ac5f2
--- /dev/null
+++ b/dev/tatoeba-dev.eng-zsm_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:377632cb44a0ea5d5ac581f8b0e41e92e4b361e4fca3d8f8220126d21241b108
+size 81188
diff --git a/dev/tatoeba-dev.epo-bos_Latn.tsv b/dev/tatoeba-dev.epo-bos_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7849f6d086105ea2a50cf00430704f38016f823d
--- /dev/null
+++ b/dev/tatoeba-dev.epo-bos_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90c3e914cb2491afda4c94626c970e158607a522ab56618f611419e28cdc8fd8
+size 1170
diff --git a/dev/tatoeba-dev.epo-cmn_Hans.tsv b/dev/tatoeba-dev.epo-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..078458bbe3533a739517cd9245a18ef2f85e1795
--- /dev/null
+++ b/dev/tatoeba-dev.epo-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c00edfa2db56dd86bc149198c56f8feb96c8ef09cd389861b2d2f5edb50bf1c0
+size 48203
diff --git a/dev/tatoeba-dev.epo-cmn_Hant.tsv b/dev/tatoeba-dev.epo-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..73cb7abac874acce70d558065bd14b69ea9b6d14
--- /dev/null
+++ b/dev/tatoeba-dev.epo-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:721c8d13a71602fd4c23adae98fb0836d1f2063bc6bf32a267067adf7a63d3db
+size 29519
diff --git a/dev/tatoeba-dev.epo-epo.tsv b/dev/tatoeba-dev.epo-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7ae3afbcad6a25a68f168772f7327647a0f17348
--- /dev/null
+++ b/dev/tatoeba-dev.epo-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b41d9e4b2c7d6034b1b3e7bea8e55f66058d9e27d07843593b70853e9342d22
+size 781069
diff --git a/dev/tatoeba-dev.epo-fas.tsv b/dev/tatoeba-dev.epo-fas.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d1a904ead5b9b308384e08c02e0ea5f0373452ce
--- /dev/null
+++ b/dev/tatoeba-dev.epo-fas.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8e346addf11dec798f6aa371fe9772e21e33bd2e474675f7d428d8c79463ff8
+size 785753
diff --git a/dev/tatoeba-dev.epo-fin.tsv b/dev/tatoeba-dev.epo-fin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4d920fad21464ead524e2b0639e0b2204c8d19ee
--- /dev/null
+++ b/dev/tatoeba-dev.epo-fin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67e5a1ed2d2457695e15c38f513b0aa7e18dd6ec171a82785831feb358106b4c
+size 75481
diff --git a/dev/tatoeba-dev.epo-fra.tsv b/dev/tatoeba-dev.epo-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d9b2239fdd1260317eb72d2b4d5df41d71de57df
--- /dev/null
+++ b/dev/tatoeba-dev.epo-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e53a041eac5099ea5f0d8682cacc8cdc308b97c972a0daaf4f589155ba95b5d
+size 25412672
diff --git a/dev/tatoeba-dev.epo-hbs.tsv b/dev/tatoeba-dev.epo-hbs.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ef579ede71e43b5aa8ba46ad35808443f39a287a
--- /dev/null
+++ b/dev/tatoeba-dev.epo-hbs.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d6d9d6963c0cde98fd8386004372c5bc64e6676ee5082aa719b292c95fc204e
+size 288686
diff --git a/dev/tatoeba-dev.epo-heb.tsv b/dev/tatoeba-dev.epo-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..98a52fad1c81a7a294feb7138150047d22e18e75
--- /dev/null
+++ b/dev/tatoeba-dev.epo-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a57deb5e715741c727187e9408100ae292086ef8ad2b39e807a6d10f14b8eb77
+size 1880897
diff --git a/dev/tatoeba-dev.epo-hrv.tsv b/dev/tatoeba-dev.epo-hrv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1600e793ab62926b967255b0e35e9c14346dec83
--- /dev/null
+++ b/dev/tatoeba-dev.epo-hrv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02b72630f34202e9b6db8ea54794f7c3e5bc22362d10e4621279a1ab58a81015
+size 35993
diff --git a/dev/tatoeba-dev.epo-hun.tsv b/dev/tatoeba-dev.epo-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1fbd26e9762631c69a0d237811cbf14b478ad0d8
--- /dev/null
+++ b/dev/tatoeba-dev.epo-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d29ba77072befe0344a8a6c3effb4bf64cab82cad716def2169ae4dcafdacace
+size 2405986
diff --git a/dev/tatoeba-dev.epo-ido.tsv b/dev/tatoeba-dev.epo-ido.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..93176e7ccf064e57ac90d91c2f2dcfbade13cb67
--- /dev/null
+++ b/dev/tatoeba-dev.epo-ido.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a16a9eed0825856b67536273aad5831db9a788147a3df9b4e3dd5ec4407c213b
+size 26630
diff --git a/dev/tatoeba-dev.epo-ile.tsv b/dev/tatoeba-dev.epo-ile.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8f211ba7d3d03e46b4bc3947f80430b34693271d
--- /dev/null
+++ b/dev/tatoeba-dev.epo-ile.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:724b4cbf0f23072911a88d060170f58b4171ecb6f174c2a0b39b27389c541912
+size 1359
diff --git a/dev/tatoeba-dev.epo-ina.tsv b/dev/tatoeba-dev.epo-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..220f92eab2598d0acae8738a9764fa8bd7a0205d
--- /dev/null
+++ b/dev/tatoeba-dev.epo-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c65a9e73d1e36241570849b70c5e2a0c3de3be5527a6c00001594f5a891bf75
+size 126157
diff --git a/dev/tatoeba-dev.epo-ita.tsv b/dev/tatoeba-dev.epo-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b43a7e2b7c9b1d587d8a5922c2768e9edb346f58
--- /dev/null
+++ b/dev/tatoeba-dev.epo-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81fd8ed9ddcd7a1abfe979c75232ae836ecfc22d2f0f9b947d46ecb019148aa5
+size 2967019
diff --git a/dev/tatoeba-dev.epo-jpn.tsv b/dev/tatoeba-dev.epo-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c948238edec675d7ad618166d7d814706fb020aa
--- /dev/null
+++ b/dev/tatoeba-dev.epo-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4efd1c5801a3fbba1405321aa0807cc5fdcdcb3eb588b7c5dfbed5263d1ff092
+size 614123
diff --git a/dev/tatoeba-dev.epo-jpn_Hani.tsv b/dev/tatoeba-dev.epo-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3043fd61ad7be3c1f9fcfc11b8cef3ebcf996ba4
--- /dev/null
+++ b/dev/tatoeba-dev.epo-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a99cbf480c0527ebaf0e8c40751d57b7b9c71e7aa2a794e0f2845ebf7723767
+size 58227
diff --git a/dev/tatoeba-dev.epo-jpn_Hira.tsv b/dev/tatoeba-dev.epo-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9ef9de4ce2803d8557074eeaa0061f0fbabbb7cb
--- /dev/null
+++ b/dev/tatoeba-dev.epo-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b54df0fce05b5230b3cd67a45a947c4f394f93cdbab7f3f41758730579dd394
+size 538946
diff --git a/dev/tatoeba-dev.epo-jpn_Kana.tsv b/dev/tatoeba-dev.epo-jpn_Kana.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4b83e5a6e7ff030c59f29868f9db07c41a15b09a
--- /dev/null
+++ b/dev/tatoeba-dev.epo-jpn_Kana.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c16ed4bcbc7f22e28dce0d865a1144c855adbf1cd9a1c46263c884a9c5ea6a2
+size 16305
diff --git a/dev/tatoeba-dev.epo-lad.tsv b/dev/tatoeba-dev.epo-lad.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f6c324e6356eb75d640a9b099377e692e6019f5d
--- /dev/null
+++ b/dev/tatoeba-dev.epo-lad.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a659d273984bc2482b467e4baf74d1ba5046ad16372076cce54f010bf535b16
+size 3567
diff --git a/dev/tatoeba-dev.epo-lad_Latn.tsv b/dev/tatoeba-dev.epo-lad_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fe462239d6acc5d4b9b0a146b33bf45588d650cc
--- /dev/null
+++ b/dev/tatoeba-dev.epo-lad_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8c2e6387f1f537cb9868240a2c517ff8ad61fce7a90cd7fc812d1cbaca41b91
+size 1855
diff --git a/dev/tatoeba-dev.epo-lat.tsv b/dev/tatoeba-dev.epo-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f0c59b9dc4510c25ed9a7144445d8cea14ca88d3
--- /dev/null
+++ b/dev/tatoeba-dev.epo-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86538353377f818c37da676b9f4b692722ecf3bc59c8936b3fbebae4b31558b4
+size 885465
diff --git a/dev/tatoeba-dev.epo-lfn.tsv b/dev/tatoeba-dev.epo-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0ba1673c0c558eb2f8e7f32d4175a702228246d0
--- /dev/null
+++ b/dev/tatoeba-dev.epo-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c0325db2cc4959affaf655050980e7bff18492f3ff04c44a1d391aad7ef8fc6
+size 22254
diff --git a/dev/tatoeba-dev.epo-lfn_Cyrl.tsv b/dev/tatoeba-dev.epo-lfn_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0dcf39a19cba549eb0f47c79df332db5fe37ea02
--- /dev/null
+++ b/dev/tatoeba-dev.epo-lfn_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5dece3f2c930f970b279156de53447668935ea5835d926d953d95e0eb35ec35
+size 7165
diff --git a/dev/tatoeba-dev.epo-lfn_Latn.tsv b/dev/tatoeba-dev.epo-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..07824304d7a2d25a8fdfa03248b6aac1db809a9c
--- /dev/null
+++ b/dev/tatoeba-dev.epo-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca6e99925b737465e1f7acd95f11dffeddbc4fee9bfa349e1f0adc8deae394a2
+size 15089
diff --git a/dev/tatoeba-dev.epo-lit.tsv b/dev/tatoeba-dev.epo-lit.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2a3748e9bfa966c1c363953d12f240bf9438ed7f
--- /dev/null
+++ b/dev/tatoeba-dev.epo-lit.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac2421f4466dea29d800a898e8a37bc962bf0bbce11b8b40e47176ef93c0969f
+size 772441
diff --git a/dev/tatoeba-dev.epo-nds.tsv b/dev/tatoeba-dev.epo-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..98bc24899ad9b3a6fad650c4a68abc345315ee64
--- /dev/null
+++ b/dev/tatoeba-dev.epo-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d9ac9fa89ddacf2e106504408c441a94c1070d4475e99d5c56d2092abb06030
+size 86612
diff --git a/dev/tatoeba-dev.epo-nld.tsv b/dev/tatoeba-dev.epo-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e1c4b42feaa5d20320949369b2c7940d2dbb3048
--- /dev/null
+++ b/dev/tatoeba-dev.epo-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60bd3fa177d8b230c2244aedba5cf3c7f5a41f5581c4be99ba3e5907de1f8ad4
+size 6484305
diff --git a/dev/tatoeba-dev.epo-nno.tsv b/dev/tatoeba-dev.epo-nno.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..814c944cb29eb6a57b76b8121a0566415055a96f
--- /dev/null
+++ b/dev/tatoeba-dev.epo-nno.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:258dde7c92786fef0813e2dcba1cea2a0c3988f427a66796c82e5129d450b022
+size 1709
diff --git a/dev/tatoeba-dev.epo-nob.tsv b/dev/tatoeba-dev.epo-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d8035cfbad4ba59df5a54c120856ee53e04d1fc8
--- /dev/null
+++ b/dev/tatoeba-dev.epo-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f51f0af30e8d7ae38903dcae5fc459f2d9d9f37426504d338d944ca36ce88848
+size 98483
diff --git a/dev/tatoeba-dev.epo-nor.tsv b/dev/tatoeba-dev.epo-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..55e35e345b7171c2ea986df314d2c7a5c7cbb911
--- /dev/null
+++ b/dev/tatoeba-dev.epo-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3cc97fda7c2c9eda31d54bdc3158888c9d2db8a3f8af32c580948992375bfbb0
+size 100192
diff --git a/dev/tatoeba-dev.epo-pol.tsv b/dev/tatoeba-dev.epo-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dd4ed80a611cee3d84df3a3e4bb0b85e45790436
--- /dev/null
+++ b/dev/tatoeba-dev.epo-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8febfef54956e3e84c55f8af8802dcdbc3d0fbdfa690a2ea1c35f6f4ccb3ebfb
+size 439099
diff --git a/dev/tatoeba-dev.epo-por.tsv b/dev/tatoeba-dev.epo-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b04bd1264b8ee5c0304938dd41353e9c949363b4
--- /dev/null
+++ b/dev/tatoeba-dev.epo-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89e321f90d9010282f90d2829d2759bbf68386e719f99eccd8d8f7bc6ddb4e78
+size 4327242
diff --git a/dev/tatoeba-dev.epo-ron.tsv b/dev/tatoeba-dev.epo-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e54e8b224bba31439b37e61198e2087c4830566a
--- /dev/null
+++ b/dev/tatoeba-dev.epo-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83005eaeb8fa9a1515e1db37cd2b5de0e7abf07e10e637e56205c2ea74277d70
+size 95906
diff --git a/dev/tatoeba-dev.epo-rus.tsv b/dev/tatoeba-dev.epo-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f1f8576e63763084a3feed8fc893c48b69f0f639
--- /dev/null
+++ b/dev/tatoeba-dev.epo-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79df0f1f3a83ae7f7cf20ee757bf0cd266e2ea50d6608c2f860e25578727629b
+size 4916871
diff --git a/dev/tatoeba-dev.epo-spa.tsv b/dev/tatoeba-dev.epo-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5e147112067fb7570bf1daf79e2793b41967b482
--- /dev/null
+++ b/dev/tatoeba-dev.epo-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1f105d3100e3e44389027ca921372f5a73958fcd835eddf14eb107378d0c6e2
+size 5443363
diff --git a/dev/tatoeba-dev.epo-srp_Cyrl.tsv b/dev/tatoeba-dev.epo-srp_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2d639dd9e466a1ffb7557712938f2999b661c705
--- /dev/null
+++ b/dev/tatoeba-dev.epo-srp_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c96832368ebb5315223e34a7bcb9d504e0c11b3d4d6a3fb12ba42b1a2dc64466
+size 229973
diff --git a/dev/tatoeba-dev.epo-srp_Latn.tsv b/dev/tatoeba-dev.epo-srp_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4092f3f824d938ba9fe25102b9337bb5962b9128
--- /dev/null
+++ b/dev/tatoeba-dev.epo-srp_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:caa514f68a47bacf7f1ca52d49959b7cc4263e29fff09ab9848bfe268f42fadb
+size 21550
diff --git a/dev/tatoeba-dev.epo-swe.tsv b/dev/tatoeba-dev.epo-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ae3cdef294ad3ab1ec602962afd23574be4a1e40
--- /dev/null
+++ b/dev/tatoeba-dev.epo-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f635fd15667521c35e019c4bcf79c2aa982d58db86a6231db20c3f2a10ac3684
+size 5259
diff --git a/dev/tatoeba-dev.epo-tlh.tsv b/dev/tatoeba-dev.epo-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..95f20532f3052486f3290e0891476a746eea05ac
--- /dev/null
+++ b/dev/tatoeba-dev.epo-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4544ded9f9e1378f0601ec44945fc81e6313c7567fcb31974de7c3dec8dbdc5
+size 9601
diff --git a/dev/tatoeba-dev.epo-toki.tsv b/dev/tatoeba-dev.epo-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9d753be8db347fb1646fe140bae41acbeee4fefb
--- /dev/null
+++ b/dev/tatoeba-dev.epo-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e85f3cf181cfba38c2f4ff54baf0db3623505659e9fd8183d59e31a54f12d3a
+size 111539
diff --git a/dev/tatoeba-dev.epo-tur.tsv b/dev/tatoeba-dev.epo-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..460f795e4682a8a7ff2aa91997e9fbdc9ca7bf17
--- /dev/null
+++ b/dev/tatoeba-dev.epo-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b0f091350d42ba54b68b6b7b61c5cf5088493c386c12679279603e2b85da3ea
+size 552160
diff --git a/dev/tatoeba-dev.epo-ukr.tsv b/dev/tatoeba-dev.epo-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fd2d7758ad914c488375fd5cdda3802aa7ade65a
--- /dev/null
+++ b/dev/tatoeba-dev.epo-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e04d8ff385eb595a332974d78db246aa893214e7b382ce50341d28f1e676e28
+size 369368
diff --git a/dev/tatoeba-dev.epo-vol.tsv b/dev/tatoeba-dev.epo-vol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c012e1d57f3fe2a849c32b884b5cafc4af6d2021
--- /dev/null
+++ b/dev/tatoeba-dev.epo-vol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3f7f2e4bfad9e8c28ada187c8dca4f2ab267eac6e3a3060a21e6acd9b6e6287
+size 2338
diff --git a/dev/tatoeba-dev.epo-yid.tsv b/dev/tatoeba-dev.epo-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..33e12ae6de769dce3d357fcbbfe48d349cb3463d
--- /dev/null
+++ b/dev/tatoeba-dev.epo-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c2c53e6ffdf82cc0563a62125c0930f2cce22cdfa8315c4de62881c71e1b080
+size 196929
diff --git a/dev/tatoeba-dev.epo-zho.tsv b/dev/tatoeba-dev.epo-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ec9705b0c2834bcd6bd717bbfad97db64fbeaa9e
--- /dev/null
+++ b/dev/tatoeba-dev.epo-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0601ec963a68af265adea8ea4dc5af55bf633bececb49b2c9844a3645bab35f7
+size 77989
diff --git a/dev/tatoeba-dev.eus-spa.tsv b/dev/tatoeba-dev.eus-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ba0400e45eae1f468baadeee28f5f9ffbd8d500e
--- /dev/null
+++ b/dev/tatoeba-dev.eus-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67c267b5058e7851290afdeb2c475130b91fc6bb4ca4ded1b4f22bcc11b2c13b
+size 73650
diff --git a/dev/tatoeba-dev.fin-fin.tsv b/dev/tatoeba-dev.fin-fin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9f85f4b53566dde0abc793653e00b6732ac9586c
--- /dev/null
+++ b/dev/tatoeba-dev.fin-fin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a78d3f91d8fbf2eb0a5cb25d994573b30bea6ce514f26a071be305db765143e2
+size 24189
diff --git a/dev/tatoeba-dev.fin-fkv.tsv b/dev/tatoeba-dev.fin-fkv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..62605653c270595ab5c151f5ffbe4eba1f0b0a44
--- /dev/null
+++ b/dev/tatoeba-dev.fin-fkv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e39a3db93ca3fa75a164e6129794dd5c99a8a607f265eb431472890980e9aec6
+size 5280
diff --git a/dev/tatoeba-dev.fin-fra.tsv b/dev/tatoeba-dev.fin-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6dc2034c90de2af93c9cf0102da2c88c8d4d5549
--- /dev/null
+++ b/dev/tatoeba-dev.fin-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a712f22a7010c6fc96e3b19dfbfa2719524b35025d5654fcda1f9d04885700c7
+size 71880
diff --git a/dev/tatoeba-dev.fin-ita.tsv b/dev/tatoeba-dev.fin-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c1f9d07a1d01e5be308b5f96ae4ded3bcdde15f8
--- /dev/null
+++ b/dev/tatoeba-dev.fin-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18e9c220ef96d243ca4a5b5edc21bea655efae39040880a9265d139038af4872
+size 69745
diff --git a/dev/tatoeba-dev.fin-jpn.tsv b/dev/tatoeba-dev.fin-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6dbf43928714f5b8d77ea9f15ec35126a2b3c5f1
--- /dev/null
+++ b/dev/tatoeba-dev.fin-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a255adcf2f2f4331b9d060cf25f2a6b9df2cdd8b457474d8fb98b2eaa197868
+size 1038239
diff --git a/dev/tatoeba-dev.fin-jpn_Hani.tsv b/dev/tatoeba-dev.fin-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..555f5c3291148dd4afcefd233ae25800fe637aa6
--- /dev/null
+++ b/dev/tatoeba-dev.fin-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6f9fa6e3e669ed1ac062c5ac959eb84d370f1cfc63da29b65d7632ebe132da7
+size 90795
diff --git a/dev/tatoeba-dev.fin-jpn_Hira.tsv b/dev/tatoeba-dev.fin-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..254c60e703e281f910ee05797d67653d67b721df
--- /dev/null
+++ b/dev/tatoeba-dev.fin-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3df03c4b4b1118045a21ecd4cf1e8f43d3c7b0c4848f1618682864aaacf95031
+size 913048
diff --git a/dev/tatoeba-dev.fin-jpn_Kana.tsv b/dev/tatoeba-dev.fin-jpn_Kana.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..51f883f160da1adc7d552e0cad1cfebd9773946c
--- /dev/null
+++ b/dev/tatoeba-dev.fin-jpn_Kana.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64f4233a12a6945cbd018f5492d810a611bcb7ff7986acb965c37ccb983ef662
+size 32000
diff --git a/dev/tatoeba-dev.fin-nno.tsv b/dev/tatoeba-dev.fin-nno.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..04675915e9283fa4ea113b9059ccc4cabb677f45
--- /dev/null
+++ b/dev/tatoeba-dev.fin-nno.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4da9a3ea4f5ad93da582da1b697eb68f279c43c32da89500111ebc2e7231454c
+size 26221
diff --git a/dev/tatoeba-dev.fin-nob.tsv b/dev/tatoeba-dev.fin-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c9d46c37e99d2ec7bf128ae5f7ce6471f6685f93
--- /dev/null
+++ b/dev/tatoeba-dev.fin-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e627dc2dd4ea31091fc08c93d504f8e8c659a52b405675cb708316abd354119f
+size 49538
diff --git a/dev/tatoeba-dev.fin-nor.tsv b/dev/tatoeba-dev.fin-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bd23afc9e1fc407347138ca217575f33ed212cd8
--- /dev/null
+++ b/dev/tatoeba-dev.fin-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be712cec3d93fa4d86406b359785860027d4f788d929dd4363086c53d9ee926e
+size 75759
diff --git a/dev/tatoeba-dev.fin-rus.tsv b/dev/tatoeba-dev.fin-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cc1a008209afc79cab3f1f5d0ffbd72a704d2938
--- /dev/null
+++ b/dev/tatoeba-dev.fin-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8aa446bc0ddd2f542acc9d5e471cadef4693b882a936883ef9cefe795aba72d
+size 91016
diff --git a/dev/tatoeba-dev.fin-spa.tsv b/dev/tatoeba-dev.fin-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a54fc28c53e5bb37231efee93a0044a08dde988e
--- /dev/null
+++ b/dev/tatoeba-dev.fin-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9829467057318f4b41a9f2d3f2154e08c010c06a0920e57df186e08b8921183b
+size 550681
diff --git a/dev/tatoeba-dev.fin-swe.tsv b/dev/tatoeba-dev.fin-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4ce572c194e32b2219b764636bfbaea726c6e6c2
--- /dev/null
+++ b/dev/tatoeba-dev.fin-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f56382fd02ee945aa8e517c45d5720367c92fbfe5dfeb7f920b07e0cc76e5270
+size 523109
diff --git a/dev/tatoeba-dev.fra-cmn_Hans.tsv b/dev/tatoeba-dev.fra-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..df48b92dd701db30e82acdea9f7a8cec834d86ac
--- /dev/null
+++ b/dev/tatoeba-dev.fra-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d9b1f420305b91c742077024cfb668742cd2d294da57639ea84a9f73c057f9e
+size 478104
diff --git a/dev/tatoeba-dev.fra-cmn_Hant.tsv b/dev/tatoeba-dev.fra-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8a14bbedf3fe61b6d00b1321f8ca06f8eb44bd5f
--- /dev/null
+++ b/dev/tatoeba-dev.fra-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae0f995684ee799959596ce0724e0e3b0321394d09014d8444b4faee9d4a35e1
+size 139241
diff --git a/dev/tatoeba-dev.fra-fra.tsv b/dev/tatoeba-dev.fra-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..941d09dd2b9398b0af3a1c1c8d7083b70703814d
--- /dev/null
+++ b/dev/tatoeba-dev.fra-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d76760ee202b0a976e6d7bf16e116fcb3fa0f8344b5da161ec0294cd2917ab02
+size 174853
diff --git a/dev/tatoeba-dev.fra-heb.tsv b/dev/tatoeba-dev.fra-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f98e9ecd7bd30322ec2631c6bf20f254d2c0772c
--- /dev/null
+++ b/dev/tatoeba-dev.fra-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c47cd616da42ba262c0189441c8ad9ee382e1a21320ec06979dc3954716db737
+size 100195
diff --git a/dev/tatoeba-dev.fra-hun.tsv b/dev/tatoeba-dev.fra-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1705d16592fbb2834ef3d9e42d39fd9a93f34822
--- /dev/null
+++ b/dev/tatoeba-dev.fra-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:790690ada71db9d790406283e161f242cd18d4222388997189157571e22d9539
+size 345870
diff --git a/dev/tatoeba-dev.fra-ido.tsv b/dev/tatoeba-dev.fra-ido.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c394eddd9609ea81a73730cb8351b754903a7354
--- /dev/null
+++ b/dev/tatoeba-dev.fra-ido.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bc017df39aa64fd61f1519bd1dba7b0785ecb1877f6db3ac76aa11d97a62215
+size 1726
diff --git a/dev/tatoeba-dev.fra-ina.tsv b/dev/tatoeba-dev.fra-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..338eb75046ee87d66261b0bb7db4887e3cec6898
--- /dev/null
+++ b/dev/tatoeba-dev.fra-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49e336a91f6d40b226ace824932aa91b85d99bce3f85398b5d7d85590e91a3c6
+size 5818
diff --git a/dev/tatoeba-dev.fra-ita.tsv b/dev/tatoeba-dev.fra-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c89d9ecf77551c411ce49ea360619f1bf1c2044b
--- /dev/null
+++ b/dev/tatoeba-dev.fra-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff310236bae2b1843d1b1d53f4b2e84270144a63cd0c8efb6f4fca5131a1f44a
+size 4894437
diff --git a/dev/tatoeba-dev.fra-jpn.tsv b/dev/tatoeba-dev.fra-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8c3fc9fa32208fdfba00fdda4c7fb1e339d4e180
--- /dev/null
+++ b/dev/tatoeba-dev.fra-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12c6e542da6a3be82bad7a75906a6b8c4a38a3ec0c7f7bb7e13e601e1b1798fd
+size 3095807
diff --git a/dev/tatoeba-dev.fra-jpn_Hani.tsv b/dev/tatoeba-dev.fra-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2ac5bc859c5186f64cd1ff99489b505c61a2acfb
--- /dev/null
+++ b/dev/tatoeba-dev.fra-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac6de2aaf87c0e303cef3300e211f31024807644782e8a8e64fc866e53c5d37f
+size 324946
diff --git a/dev/tatoeba-dev.fra-jpn_Hira.tsv b/dev/tatoeba-dev.fra-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c1fa87d4b281b3eb8e849f40e1e1d1df491e1ea7
--- /dev/null
+++ b/dev/tatoeba-dev.fra-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c01dcde51f46b9a388e9ea0b602a4c40de9e760eaad1e842ee35f1df5ebb273
+size 2719312
diff --git a/dev/tatoeba-dev.fra-jpn_Kana.tsv b/dev/tatoeba-dev.fra-jpn_Kana.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c008342b63b2d701c32311a0be602854d7aa13db
--- /dev/null
+++ b/dev/tatoeba-dev.fra-jpn_Kana.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e66202b4bb9af242168e8eee187ab0604b2fa85312f55d8b3b642fdd344f177f
+size 49707
diff --git a/dev/tatoeba-dev.fra-kab.tsv b/dev/tatoeba-dev.fra-kab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0390e097627ed9425f1e93dca36038521a3ab058
--- /dev/null
+++ b/dev/tatoeba-dev.fra-kab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0efd36888a742908198eda972dd6f37e7134938af502cf1d73ce2373fdf63589
+size 1511811
diff --git a/dev/tatoeba-dev.fra-lat.tsv b/dev/tatoeba-dev.fra-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..38eed2919c2ebd4a2b857f51043cca1c9d839318
--- /dev/null
+++ b/dev/tatoeba-dev.fra-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:613d16a2d27c3354d193996658e9d95912e6b89b170d0263ae1a98bd5481f95f
+size 137312
diff --git a/dev/tatoeba-dev.fra-lfn.tsv b/dev/tatoeba-dev.fra-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a9f3a95f86fe63e5dff6a5dc5c0da9d01267f6a4
--- /dev/null
+++ b/dev/tatoeba-dev.fra-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:823b157064197427a003bf7b9617b96d243165f996eec84b52663bc6e4348eeb
+size 16375
diff --git a/dev/tatoeba-dev.fra-lfn_Cyrl.tsv b/dev/tatoeba-dev.fra-lfn_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3c1e4941570f2c84c3a739343b80e4cfd5fe4eef
--- /dev/null
+++ b/dev/tatoeba-dev.fra-lfn_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a06052dfccd4875871aa98a2ec8a1f9915534d0c37e4f9ddda3987479acdeec
+size 9692
diff --git a/dev/tatoeba-dev.fra-lfn_Latn.tsv b/dev/tatoeba-dev.fra-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c5dac067ffcbcc781fbd28f709bf6e4d7a46b943
--- /dev/null
+++ b/dev/tatoeba-dev.fra-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9dab178f827e6606a3b9f2d668c7a84669a899669e81aa07a57c3cfaebb45661
+size 6683
diff --git a/dev/tatoeba-dev.fra-lzh_Hani.tsv b/dev/tatoeba-dev.fra-lzh_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b07e6afc78d8913f6aeaac8a56796c6ba70779e8
--- /dev/null
+++ b/dev/tatoeba-dev.fra-lzh_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9214edc446f749bca4ea52629e04fc562c9c02c38368880604331eaadfc41695
+size 2291
diff --git a/dev/tatoeba-dev.fra-nld.tsv b/dev/tatoeba-dev.fra-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9844ee1b8799e8267b092caef283ccad1c481a8e
--- /dev/null
+++ b/dev/tatoeba-dev.fra-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed4a30e167d5b88e7cfc9673ad925f1357835957d4449c1ad8f0337bece3b07e
+size 1233531
diff --git a/dev/tatoeba-dev.fra-pol.tsv b/dev/tatoeba-dev.fra-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..eef2cea82f703d191306374f6b1b037ae0c31453
--- /dev/null
+++ b/dev/tatoeba-dev.fra-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0122db9dbfbe8c18d06fdd06264fc5fb9c48e3cfb9faa86203faf0452f2f81c2
+size 85755
diff --git a/dev/tatoeba-dev.fra-por.tsv b/dev/tatoeba-dev.fra-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..88f7bcdcf94002178022b2fad42e30237935167b
--- /dev/null
+++ b/dev/tatoeba-dev.fra-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f72e402546d25eaac55e9fdb102e9b03489276995adc5c0a98e9a0a362c42931
+size 1353409
diff --git a/dev/tatoeba-dev.fra-rus.tsv b/dev/tatoeba-dev.fra-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7992afe0622ca39f2c934e6fafdc17cbe0d7a9ff
--- /dev/null
+++ b/dev/tatoeba-dev.fra-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1462f8ae27b4d79f3706e07b792d4030ead57932cfeffe19938f9e6a3865c20e
+size 17211453
diff --git a/dev/tatoeba-dev.fra-spa.tsv b/dev/tatoeba-dev.fra-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7402da3f6e2ce67154684193f1ca6aea1604cd9a
--- /dev/null
+++ b/dev/tatoeba-dev.fra-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3539fe6b0f20c5f0e837571a4fe0fa934935b047e44688dfd80aa0dc565b7349
+size 3483005
diff --git a/dev/tatoeba-dev.fra-swe.tsv b/dev/tatoeba-dev.fra-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e7894732dabde7a97e3ddd9e4447fe82fe4ef9bf
--- /dev/null
+++ b/dev/tatoeba-dev.fra-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:269a1770ad97cde4877be7272c0e8b860524e066bb01cf9e63f9231ae57a7873
+size 79547
diff --git a/dev/tatoeba-dev.fra-toki.tsv b/dev/tatoeba-dev.fra-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7311ba28aa7c37ef71ccea24145db4d3debfe81d
--- /dev/null
+++ b/dev/tatoeba-dev.fra-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e5807fb920ce52635f2cd44e6f174c4390a56485736ffc994712fb9dcf23ad5
+size 3337
diff --git a/dev/tatoeba-dev.fra-tur.tsv b/dev/tatoeba-dev.fra-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e1427ccb03d41e47e5911d4ca8dc134b9100d98b
--- /dev/null
+++ b/dev/tatoeba-dev.fra-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35be02b27797f30a08bcf2cabfd5b2dcac445be46081468c3c6982889b1a26e9
+size 543390
diff --git a/dev/tatoeba-dev.fra-ukr.tsv b/dev/tatoeba-dev.fra-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a58c9631aa407788c3c256565cf4b53736a4bcd1
--- /dev/null
+++ b/dev/tatoeba-dev.fra-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d6ca4fddd2d41d6be0fe41d6af277796946b82329011781e6fbc4577529a2cd
+size 1522132
diff --git a/dev/tatoeba-dev.fra-wuu.tsv b/dev/tatoeba-dev.fra-wuu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3f8e990144b6cc0d5521f9bf19162fefd8c87e2d
--- /dev/null
+++ b/dev/tatoeba-dev.fra-wuu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40d78de13b40aa20b63f9ab25c6e7df4eb471946ea42cf4348672794214cf3bd
+size 41520
diff --git a/dev/tatoeba-dev.fra-wuu_Hani.tsv b/dev/tatoeba-dev.fra-wuu_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cb20f2bb1241fa01ea162d7e03de07a7048245f6
--- /dev/null
+++ b/dev/tatoeba-dev.fra-wuu_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba6e0f739d9d83851a9862a4dc4cdf7012ec9c99acca1fbd3e6bd41cb5cbafc8
+size 2863
diff --git a/dev/tatoeba-dev.fra-yid.tsv b/dev/tatoeba-dev.fra-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..347af109ffef7af41b24808842a17acdbd8a835e
--- /dev/null
+++ b/dev/tatoeba-dev.fra-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adb9714b964368b2def30f552fdd0f8d1491c804dea4c89110613ecd10f1d7f3
+size 6092
diff --git a/dev/tatoeba-dev.fra-yue_Hans.tsv b/dev/tatoeba-dev.fra-yue_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..731cfcf9feee8cd8ad317358f39db489b1af6d3b
--- /dev/null
+++ b/dev/tatoeba-dev.fra-yue_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9eaffc3c33cb507d6f92fa217c0e3b1208a6b149d648a20c13d5558a043e72be
+size 6198
diff --git a/dev/tatoeba-dev.fra-yue_Hant.tsv b/dev/tatoeba-dev.fra-yue_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..191a5931da3adb9dfddba04829f264ba1c45da69
--- /dev/null
+++ b/dev/tatoeba-dev.fra-yue_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ed1e4c0f140281e72ad3d5c570295d817f818fbc6b5f0394ffed0601bdd4480
+size 3515
diff --git a/dev/tatoeba-dev.fra-zho.tsv b/dev/tatoeba-dev.fra-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..436a384f12d7a49b3957d7b88122476a16e37bfe
--- /dev/null
+++ b/dev/tatoeba-dev.fra-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38d26d970053e0bd5b4908037d1f3e8259dd17518b7e4a2634b7edd4ea3f9c48
+size 674827
diff --git a/dev/tatoeba-dev.glg-spa.tsv b/dev/tatoeba-dev.glg-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d8d35761783993478a878c9f9dab42dadd58ccc9
--- /dev/null
+++ b/dev/tatoeba-dev.glg-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e65e077a1fbac24e376479c58656b0482f50e690506c0fb9450fa357bbe0f6d5
+size 86358
diff --git a/dev/tatoeba-dev.gos-nld.tsv b/dev/tatoeba-dev.gos-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5a16d9b6ace0395a55a952fb0e07dd97abef278f
--- /dev/null
+++ b/dev/tatoeba-dev.gos-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:411299ede82e78e45e3bef95778f544d658815e2292e1fa63c52fbba90e55187
+size 26165
diff --git a/dev/tatoeba-dev.grn-por.tsv b/dev/tatoeba-dev.grn-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3b5445f042af5a9a173434235cf0721c948335bc
--- /dev/null
+++ b/dev/tatoeba-dev.grn-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35cf6efaf42d87e233e1c7d6cfb45ce3c34b36eac54e3843b89bc610b2e52673
+size 5806
diff --git a/dev/tatoeba-dev.grn-spa.tsv b/dev/tatoeba-dev.grn-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..58b876d08ec9b1afb394160947eaa7b4ff1dcf58
--- /dev/null
+++ b/dev/tatoeba-dev.grn-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c00f15bb7e51bf083fbda19cc17854b315a07b6c271c56a4275099d221372d45
+size 11168
diff --git a/dev/tatoeba-dev.hbs-nor.tsv b/dev/tatoeba-dev.hbs-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f3efafccec48f7fc7835a2c107565fd4f24e734c
--- /dev/null
+++ b/dev/tatoeba-dev.hbs-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61dc6897428ca7b1463448815127565f135d437639776c0359e800dfcff9996d
+size 575038
diff --git a/dev/tatoeba-dev.hbs-rus.tsv b/dev/tatoeba-dev.hbs-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0390d879679b3175b121dc92aa80951ed3ea16a9
--- /dev/null
+++ b/dev/tatoeba-dev.hbs-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57497052e392e2ef5d3d940d40d4b9cb8956effde82669f2116fe45bd21a11e2
+size 397416
diff --git a/dev/tatoeba-dev.heb-heb.tsv b/dev/tatoeba-dev.heb-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ba215af5eb9a9facec0e2fe5f462a5516f2e8c30
--- /dev/null
+++ b/dev/tatoeba-dev.heb-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25c5d8641d11735f2bd84755994b2c199af8ac9729aecc8d670b70f34acdf599
+size 71416
diff --git a/dev/tatoeba-dev.heb-ina.tsv b/dev/tatoeba-dev.heb-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f3679eae0a7a9337dc72660fd456c439b1a47c2a
--- /dev/null
+++ b/dev/tatoeba-dev.heb-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e982d153261b194f2311195fd6205fbb36b0152b2846dd1e76c6e80fb17cb4f
+size 3346
diff --git a/dev/tatoeba-dev.heb-ita.tsv b/dev/tatoeba-dev.heb-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8c9500324e4af8b04dedf141e87ddee5bcc54589
--- /dev/null
+++ b/dev/tatoeba-dev.heb-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04f2bb2a052cc39f28d427117b5a70a1f9d317653f4acf795a42c93972500b2a
+size 2202
diff --git a/dev/tatoeba-dev.heb-lad.tsv b/dev/tatoeba-dev.heb-lad.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..52d08ed922428b90f4f8544e53af9ae7a26d4142
--- /dev/null
+++ b/dev/tatoeba-dev.heb-lad.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2405a45e5cb31db213d953040d5e2730630aa12f7d84a9f6c832b2239c711c74
+size 3126
diff --git a/dev/tatoeba-dev.heb-lad_Latn.tsv b/dev/tatoeba-dev.heb-lad_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..628e4c42eb71e9088dc82e009f0897544b9927b5
--- /dev/null
+++ b/dev/tatoeba-dev.heb-lad_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:764e3000e43734f4a18f6639bce384f5270df978b0a924b981e7276de0fde1d1
+size 1549
diff --git a/dev/tatoeba-dev.heb-lat.tsv b/dev/tatoeba-dev.heb-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9c6cfc06d9319d2216c3127404ac9e6c0a455966
--- /dev/null
+++ b/dev/tatoeba-dev.heb-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb327eeba06ecf9ac563896cbcb0767c244c935a61ae7681c55463f5c84b41b7
+size 4929
diff --git a/dev/tatoeba-dev.heb-lfn.tsv b/dev/tatoeba-dev.heb-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fc385bc73d0d5c344035b31de35a497102d24b06
--- /dev/null
+++ b/dev/tatoeba-dev.heb-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:340e81c3116fdd833cf9b7a68bbda995e451c6a8c981f6337b50e8b9813a265c
+size 6388
diff --git a/dev/tatoeba-dev.heb-lfn_Cyrl.tsv b/dev/tatoeba-dev.heb-lfn_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ad3b1e914e960bae54f31a3e4b89c7a95a501b30
--- /dev/null
+++ b/dev/tatoeba-dev.heb-lfn_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bd2d13c050d7b3132d72b2f57e99840899243bd347a39aa8984dcfe09c464d4
+size 2966
diff --git a/dev/tatoeba-dev.heb-lfn_Latn.tsv b/dev/tatoeba-dev.heb-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ed22a8018c7a29763bb0fce05a9c0671d93218b6
--- /dev/null
+++ b/dev/tatoeba-dev.heb-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:912a6ff3f261b0de77f862ce7ba5baf425883353b86da8f285d5b2279151d636
+size 3422
diff --git a/dev/tatoeba-dev.heb-nld.tsv b/dev/tatoeba-dev.heb-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1fbc9c61e2c3ebd5e4776cedceff126b640edd84
--- /dev/null
+++ b/dev/tatoeba-dev.heb-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f35532008c4c89dcf03695f12881cfe6f0d4548aad9492502aa1ee1bc04d577a
+size 416293
diff --git a/dev/tatoeba-dev.heb-pol.tsv b/dev/tatoeba-dev.heb-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f671b4840403700224385124feec041f35faa009
--- /dev/null
+++ b/dev/tatoeba-dev.heb-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0100aab812c0532636e8961d7debfd38cc73a29a4f074fa74c61af7c42235ff
+size 682085
diff --git a/dev/tatoeba-dev.heb-por.tsv b/dev/tatoeba-dev.heb-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d84b65a277db4c4aa0c7dfe2fbdd896bfb222847
--- /dev/null
+++ b/dev/tatoeba-dev.heb-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59032ac9e2b7ecccc34d5408afad03e25e84ada707815cf356a7addcb679288c
+size 1554
diff --git a/dev/tatoeba-dev.heb-rus.tsv b/dev/tatoeba-dev.heb-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1b7bb531badf8019847629df30349eb0a7d81b18
--- /dev/null
+++ b/dev/tatoeba-dev.heb-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf7a4f71c5e321e97498f7d34304f6ccce9452591c072639ed6be673cc2ff097
+size 386010
diff --git a/dev/tatoeba-dev.heb-spa.tsv b/dev/tatoeba-dev.heb-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..21bf343bbe77b7f55143571e5726adae3d43f4db
--- /dev/null
+++ b/dev/tatoeba-dev.heb-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84b173df9b06b4054bdd5b41d7de949b98153ae1da546ece8a9b0151318f570c
+size 99049
diff --git a/dev/tatoeba-dev.heb-tur.tsv b/dev/tatoeba-dev.heb-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..03b284df3f08c0e1407ee9b38ec463086f4a8b44
--- /dev/null
+++ b/dev/tatoeba-dev.heb-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a0bb787131d2858bdb92ed1bffbdf87e75b7eca0ecbfeadb785ba7e0ccb0c0a
+size 1985
diff --git a/dev/tatoeba-dev.heb-yid.tsv b/dev/tatoeba-dev.heb-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..be9c40207e25a9fa3857bff6bdb66a48ca4794e2
--- /dev/null
+++ b/dev/tatoeba-dev.heb-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:579f7a0ccbfd568252659824a45480fc9acd429efa756a3f87591b651142a922
+size 21075
diff --git a/dev/tatoeba-dev.hrv-nob.tsv b/dev/tatoeba-dev.hrv-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5fce319cce8742fa18247a5775831a4f945eb515
--- /dev/null
+++ b/dev/tatoeba-dev.hrv-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bfa96c80573e41aca95830000c18b1dd22e7948301500c2c9d8da7ffca1c56dc
+size 4262
diff --git a/dev/tatoeba-dev.hrv-rus.tsv b/dev/tatoeba-dev.hrv-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9330c9e008296e4a8e11e17eece8df00150f1495
--- /dev/null
+++ b/dev/tatoeba-dev.hrv-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fab97ae43ac13b3fc3ef07f01e07a5dccdca9e562e2d2680aca360659ad4ff8d
+size 18939
diff --git a/dev/tatoeba-dev.hun-ita.tsv b/dev/tatoeba-dev.hun-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..afcfc370bd00cc2dd17193bb409d2d51a73dc345
--- /dev/null
+++ b/dev/tatoeba-dev.hun-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b00a78fa7a342f7c6e7e66520bdf91bafd58645bb299c8972bf6670331cc553
+size 468816
diff --git a/dev/tatoeba-dev.hun-jpn.tsv b/dev/tatoeba-dev.hun-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0e3985ec295a1e8bd3dcebdd422ef1ca2069cb99
--- /dev/null
+++ b/dev/tatoeba-dev.hun-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39f52cbeea00e99bada219a355b85815c57c0396036fd84cee91f94410212327
+size 409758
diff --git a/dev/tatoeba-dev.hun-jpn_Hani.tsv b/dev/tatoeba-dev.hun-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3587857c95bcd39c1a4047ca08a4e23d5121cb9b
--- /dev/null
+++ b/dev/tatoeba-dev.hun-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3352a419371d5a10387a92082b28df0f580bab340752aaaec0e42e4649882309
+size 62022
diff --git a/dev/tatoeba-dev.hun-jpn_Hira.tsv b/dev/tatoeba-dev.hun-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f7cf4276658cb393d35b6cfd32f787a0dd44f72f
--- /dev/null
+++ b/dev/tatoeba-dev.hun-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:915ec7a96ae33e0a36ae15ae762e1fab3a53139cfc6bb14b76e0214d11d56353
+size 342924
diff --git a/dev/tatoeba-dev.hun-jpn_Kana.tsv b/dev/tatoeba-dev.hun-jpn_Kana.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..30fe252e716b6d62e7ddd1f8e3ce0da38bae3acb
--- /dev/null
+++ b/dev/tatoeba-dev.hun-jpn_Kana.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:001b14700ea08acd469f2361b25c785ba1f22ef32199c5c3e6af5f5973caa8fe
+size 3812
diff --git a/dev/tatoeba-dev.hun-nld.tsv b/dev/tatoeba-dev.hun-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4ac6c7eb07b8ba29fe05c0a5ccf9d4565138a265
--- /dev/null
+++ b/dev/tatoeba-dev.hun-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a5429b57214de6dbdde07e546c4dec20728579760ac5ff0fe20446f21e83a6c
+size 2066
diff --git a/dev/tatoeba-dev.hun-por.tsv b/dev/tatoeba-dev.hun-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4bea556bcd5574b62bd62ddb5eea43a6d13df4f7
--- /dev/null
+++ b/dev/tatoeba-dev.hun-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0efb13c82899429e3f877866422ab3f99cf52b0edf724753f46424a05e6a2102
+size 248574
diff --git a/dev/tatoeba-dev.hun-rus.tsv b/dev/tatoeba-dev.hun-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e66e76dc6ebd558f265ac541ded211224f31f5b4
--- /dev/null
+++ b/dev/tatoeba-dev.hun-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0dcb9aa8a958dd5b5a9e9eab28693cf6b5630e30563128d85a0f70c10860c756
+size 565264
diff --git a/dev/tatoeba-dev.hun-spa.tsv b/dev/tatoeba-dev.hun-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8f78fcfd03dd8ee8d95fdf1b703c5fae44608e57
--- /dev/null
+++ b/dev/tatoeba-dev.hun-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fec27543afb1552e793855a48c88a31ea911f0eefa549e2f4c1677ae610adcd
+size 303461
diff --git a/dev/tatoeba-dev.hun-swe.tsv b/dev/tatoeba-dev.hun-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..68bbe39b951ccb42749c95172f9d78cb1dc24034
--- /dev/null
+++ b/dev/tatoeba-dev.hun-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73edfb7e73c49820d697da955671abd6ff74280fce4dde73acd26ac7509dc4ad
+size 35832
diff --git a/dev/tatoeba-dev.ido-ina.tsv b/dev/tatoeba-dev.ido-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..89d628cf082011aa46e8bf7d5e8735571836b81f
--- /dev/null
+++ b/dev/tatoeba-dev.ido-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1c3eb415eddc3926556fc674cd765b97c262d10634c536e52efabc6c8a13cdd
+size 3612
diff --git a/dev/tatoeba-dev.ido-ita.tsv b/dev/tatoeba-dev.ido-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6c5c9828ceba6f852848830764c90f291804d623
--- /dev/null
+++ b/dev/tatoeba-dev.ido-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7415129f0d790c64bdf8f062f8c5399f091188ceee2f8731564c6a7b69ddb2bc
+size 2423
diff --git a/dev/tatoeba-dev.ido-lfn.tsv b/dev/tatoeba-dev.ido-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a0d5591583f1ce8114f246ce004b5980e00eb94e
--- /dev/null
+++ b/dev/tatoeba-dev.ido-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14dd78ef70f2489ad4bcb65313cd3b6fc2766e9a2b613dbb66355fc21ff37e32
+size 3845
diff --git a/dev/tatoeba-dev.ido-yid.tsv b/dev/tatoeba-dev.ido-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f770248df20ae911bc3e84b47dec2a266437751f
--- /dev/null
+++ b/dev/tatoeba-dev.ido-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e0b0a01da5abbb13215f6eb999e94e16df86279facf6e1f050e432ca6a55924
+size 5495
diff --git a/dev/tatoeba-dev.ido_Latn-lfn_Latn.tsv b/dev/tatoeba-dev.ido_Latn-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..36e643a0e02cadd9e964300a94be22e947fb3cb1
--- /dev/null
+++ b/dev/tatoeba-dev.ido_Latn-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da7e2372ff4218c4b7bf5a793731e46f4a4593d564b648311166add7e30b94bf
+size 2463
diff --git a/dev/tatoeba-dev.ina-lad.tsv b/dev/tatoeba-dev.ina-lad.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0df7cdb1c4b9494acbe4a046cdff582f3aec6d3e
--- /dev/null
+++ b/dev/tatoeba-dev.ina-lad.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adb4bd0974819d920019f635f33ddee25e1239411fbef7dd324d4b983902d0bb
+size 3378
diff --git a/dev/tatoeba-dev.ina-lat.tsv b/dev/tatoeba-dev.ina-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..73e93d344b5c4ebd2067404f37a11daaeb635e48
--- /dev/null
+++ b/dev/tatoeba-dev.ina-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:620c2c42126a5757db97ac15deb460b9f39f8c26dc2825438e98395f2e88b2ed
+size 4698
diff --git a/dev/tatoeba-dev.ina-lfn.tsv b/dev/tatoeba-dev.ina-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b7053dd548018dfbf21c6ae9a0297c1d3231537e
--- /dev/null
+++ b/dev/tatoeba-dev.ina-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b91c6694c30620c4f159ac7c3a22e1181e6b62729adc9f96eff08abd0fbe7692
+size 9054
diff --git a/dev/tatoeba-dev.ina-por.tsv b/dev/tatoeba-dev.ina-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b0711075e052a0e215d3c9c5a1d7a6ad0ab7d2e9
--- /dev/null
+++ b/dev/tatoeba-dev.ina-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0004e828d9dd16af783eff27258ad69bc91330a31d137817a1f944888c7fed33
+size 783848
diff --git a/dev/tatoeba-dev.ina-rus.tsv b/dev/tatoeba-dev.ina-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b9ca6c109fb4c5a8a4f8ee0d6bfaf6cbe8a53d65
--- /dev/null
+++ b/dev/tatoeba-dev.ina-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35a7a50d73172be73dd330851871cba9e9d5a367d79f2e30af245f86e7c23a37
+size 2799
diff --git a/dev/tatoeba-dev.ina-spa.tsv b/dev/tatoeba-dev.ina-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..15af40b7fa72302c73d9a5c497a13a0d1af2aa36
--- /dev/null
+++ b/dev/tatoeba-dev.ina-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28dfc286362340c02231d299360704704907bf36fbe22abc1505d9722a7aecdc
+size 7849
diff --git a/dev/tatoeba-dev.ina-tlh.tsv b/dev/tatoeba-dev.ina-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3f8786f67bd294c66b9bcf37a961771a49b91b24
--- /dev/null
+++ b/dev/tatoeba-dev.ina-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe925ff342422725f9f2eab96c70d9e596f8b640398bf7a664b4f65c8fc61d1c
+size 2063
diff --git a/dev/tatoeba-dev.ina-yid.tsv b/dev/tatoeba-dev.ina-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e75060798bb87180591429da12ed928bcbfe01c1
--- /dev/null
+++ b/dev/tatoeba-dev.ina-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88b2d51744bdd73706fd0d3b1e857dd42b5955a8807f400c3f571691c52110da
+size 12363
diff --git a/dev/tatoeba-dev.ina_Latn-lad_Latn.tsv b/dev/tatoeba-dev.ina_Latn-lad_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..25f62a2797c6de6b116a66aac94adabaa3adea12
--- /dev/null
+++ b/dev/tatoeba-dev.ina_Latn-lad_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9dc2eda0d933c901cc700232db4582ecb7c40fbad69579569280756e6bba76d4
+size 1848
diff --git a/dev/tatoeba-dev.ina_Latn-lfn_Cyrl.tsv b/dev/tatoeba-dev.ina_Latn-lfn_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..019dde7595e212557d48de53962f2f5bddb18f14
--- /dev/null
+++ b/dev/tatoeba-dev.ina_Latn-lfn_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d25a734adbe26832863383e53fdb8da9acb2c3409e4e5cee54ab926efdeb443f
+size 2279
diff --git a/dev/tatoeba-dev.ina_Latn-lfn_Latn.tsv b/dev/tatoeba-dev.ina_Latn-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d86cae866999ba2411de34128ab934aa4fa52bd5
--- /dev/null
+++ b/dev/tatoeba-dev.ina_Latn-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b4fe3eb6815a41a9477674bc87b5d689d224743f6407b3bed74be124e99bb11
+size 6775
diff --git a/dev/tatoeba-dev.ita-cmn_Hans.tsv b/dev/tatoeba-dev.ita-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..40263423a1e03eb793cdc12e3cf0e0b480221c36
--- /dev/null
+++ b/dev/tatoeba-dev.ita-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f56092d4d2f6657ea3ef543b4a5388357d69321fe2eb9b68af0babd88418129
+size 45045
diff --git a/dev/tatoeba-dev.ita-cmn_Hant.tsv b/dev/tatoeba-dev.ita-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d540f0df5e59831b70959c33c3ce4b852a25963c
--- /dev/null
+++ b/dev/tatoeba-dev.ita-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f7d591a49572d817fbc128b68b5b22e7228d2358c3b6bed154641aa4ca7cd7a
+size 28675
diff --git a/dev/tatoeba-dev.ita-ita.tsv b/dev/tatoeba-dev.ita-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..347248c759c3fd9d693f3e8a900569622f8ce5cd
--- /dev/null
+++ b/dev/tatoeba-dev.ita-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8da26126c9f29d8e7d0a7fa6449c02a267b410e58d6847e8c63b3da000d503dd
+size 27857
diff --git a/dev/tatoeba-dev.ita-jpn.tsv b/dev/tatoeba-dev.ita-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bdafd0688eb388e2ff613a28aca27fa486aa94e5
--- /dev/null
+++ b/dev/tatoeba-dev.ita-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf5b2e527b542feb4f95b2d91461c1f88d10a4d268a69cd0c4ac6ab4fc97df9d
+size 106080
diff --git a/dev/tatoeba-dev.ita-jpn_Hani.tsv b/dev/tatoeba-dev.ita-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ac04892b7f185087febe5fba053352fe2304e09b
--- /dev/null
+++ b/dev/tatoeba-dev.ita-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f7148441a864e73d2dc25fe6ae5d9a8f04656e6ff20b37e2794cf52efc060480
+size 10247
diff --git a/dev/tatoeba-dev.ita-jpn_Hira.tsv b/dev/tatoeba-dev.ita-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d10adb1cae4207715d96b8aa88daa409f0f76c64
--- /dev/null
+++ b/dev/tatoeba-dev.ita-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8939d838a41fc3babb45fe8ac339c7e6abadfcae6dcda2400f8fb4be658058c7
+size 93495
diff --git a/dev/tatoeba-dev.ita-lat.tsv b/dev/tatoeba-dev.ita-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..218fc563a99b02b53d1af13dee70ee1b22ffc4c7
--- /dev/null
+++ b/dev/tatoeba-dev.ita-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21edd5652cff109006e12be2365aaf4a826fe3534559d7925fef4d35c32f9ad9
+size 1314
diff --git a/dev/tatoeba-dev.ita-nld.tsv b/dev/tatoeba-dev.ita-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a92500ddf2ca90a4388304363e54517b8ed193b9
--- /dev/null
+++ b/dev/tatoeba-dev.ita-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c23ed752a3b51dea246a4cb7a82286531c58867e45dbe560db80c4d9a29a5fc5
+size 469408
diff --git a/dev/tatoeba-dev.ita-pol.tsv b/dev/tatoeba-dev.ita-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..40875691fb818d589ec3a2579a6e411dccc5b172
--- /dev/null
+++ b/dev/tatoeba-dev.ita-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12b118c61282b711c276a2035b5543a4e0d814a649851a00774a3461894ef484
+size 73569
diff --git a/dev/tatoeba-dev.ita-por.tsv b/dev/tatoeba-dev.ita-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c41c35d7f1776addf89cbc7f36ed5139ad55efa7
--- /dev/null
+++ b/dev/tatoeba-dev.ita-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bf61fb46d77cf81f725784a856674cd73cea74a076c4175af77e4ccce055318
+size 513921
diff --git a/dev/tatoeba-dev.ita-rus.tsv b/dev/tatoeba-dev.ita-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9e0f12e6463d938f0b8fa69a1ace7e26f2d4927d
--- /dev/null
+++ b/dev/tatoeba-dev.ita-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94a92079121b07a7c3d383fdccce072d07ce8c5d41f5598cb0728138ccb425a0
+size 6559508
diff --git a/dev/tatoeba-dev.ita-spa.tsv b/dev/tatoeba-dev.ita-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2dacd037762d3d0ed0e122d092bfa3cdca323692
--- /dev/null
+++ b/dev/tatoeba-dev.ita-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4907a38de5fb707a9f4143d3566d9782fa6a1f1cc21fed04d5b155bd24cfaa24
+size 705745
diff --git a/dev/tatoeba-dev.ita-tur.tsv b/dev/tatoeba-dev.ita-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f3f81fdc7065e4809bf58a70a6491e24bb836cb0
--- /dev/null
+++ b/dev/tatoeba-dev.ita-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08b88787ec61e3ffd422946d7354719c50705044304adf857c74f8c4d4ac7a65
+size 507078
diff --git a/dev/tatoeba-dev.ita-ukr.tsv b/dev/tatoeba-dev.ita-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..59c0545f4c6f8100dcfbf6d520a1ee290d69594f
--- /dev/null
+++ b/dev/tatoeba-dev.ita-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5c0112c900d07295d9c1562a24f465e9c8f393566a23088c16375e5f3e96be8
+size 655445
diff --git a/dev/tatoeba-dev.ita-yid.tsv b/dev/tatoeba-dev.ita-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ff2436b713df5bcb76ac3a08fad6129499361fe9
--- /dev/null
+++ b/dev/tatoeba-dev.ita-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3297542b9cce3718ab762ee32d83cebd36b004974a8cf494498fadac2e9bcdd
+size 3609
diff --git a/dev/tatoeba-dev.ita-zho.tsv b/dev/tatoeba-dev.ita-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6c3516323a961faf02dea8341364dd4aa62160f9
--- /dev/null
+++ b/dev/tatoeba-dev.ita-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:745a86b5fc9caf34a864676ced7cc53fa65c5ff99e5f37b6753a54fe4db678f8
+size 75878
diff --git a/dev/tatoeba-dev.jbo-zho.tsv b/dev/tatoeba-dev.jbo-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c1f569abbf67d2508126b5f3c53cd18f2789e7ea
--- /dev/null
+++ b/dev/tatoeba-dev.jbo-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9269a284a5b27c4f1963f2bfc1cc7f72533591ca9d247a49d616e04a679a731d
+size 1715
diff --git a/dev/tatoeba-dev.jbo_Latn-cmn_Hant.tsv b/dev/tatoeba-dev.jbo_Latn-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0e2281c3297710b52748aee3244c53e6d4e6059d
--- /dev/null
+++ b/dev/tatoeba-dev.jbo_Latn-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b75608f4262c32a21d68a2382ad6190d615aeff5842dff7f08779a61021ddf1
+size 1549
diff --git a/dev/tatoeba-dev.jpn-msa.tsv b/dev/tatoeba-dev.jpn-msa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f363087467be74da90bb969b103b931f4391c931
--- /dev/null
+++ b/dev/tatoeba-dev.jpn-msa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8065bb7935921d8d96b56ac94ad4af3b9afe25910cc3dd362874dd6b6aaa9a9f
+size 98191
diff --git a/dev/tatoeba-dev.jpn-nds.tsv b/dev/tatoeba-dev.jpn-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3763451ac5fe00035dc886f995af014dcfd2a135
--- /dev/null
+++ b/dev/tatoeba-dev.jpn-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f950d59fdfda166c03206b041649b0f947d99abfe52446b84576bda2d80f1b15
+size 5011
diff --git a/dev/tatoeba-dev.jpn-nld.tsv b/dev/tatoeba-dev.jpn-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5138fd306a68a03127decbad110e84a85381aa31
--- /dev/null
+++ b/dev/tatoeba-dev.jpn-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d580ca0cfa93335954d5e12fc5c037148b8b232bb14df6ac692e56b15f4e1d2
+size 95717
diff --git a/dev/tatoeba-dev.jpn-pol.tsv b/dev/tatoeba-dev.jpn-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9f7bc983ad1173191371ac4952c09e6f43fa2125
--- /dev/null
+++ b/dev/tatoeba-dev.jpn-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0d0fbd648807bd4941895407c81d428f22fc40edb2708f2d5602ce1f8ea787c
+size 1553466
diff --git a/dev/tatoeba-dev.jpn-por.tsv b/dev/tatoeba-dev.jpn-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..26748f898a8ff4999a2ddea8415b2b7532c00b78
--- /dev/null
+++ b/dev/tatoeba-dev.jpn-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a28ce519cea18954bb83df6c56d7c009ccbf677068921660812c2cc4eef3dfa6
+size 113755
diff --git a/dev/tatoeba-dev.jpn-rus.tsv b/dev/tatoeba-dev.jpn-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f9918c3fcb6cc3bd934fa6c37c0bfd04fdf11c23
--- /dev/null
+++ b/dev/tatoeba-dev.jpn-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a49cbc65b95af7f16a46a792835f35e1d45876435a8a61a4329a8377b79c0044
+size 1946664
diff --git a/dev/tatoeba-dev.jpn-spa.tsv b/dev/tatoeba-dev.jpn-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e8662f2113de197a33c432163a38c30d53ba2ad6
--- /dev/null
+++ b/dev/tatoeba-dev.jpn-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48476680c79091388e667067216a21b52124cad5a0fd54b211714105e8af0f5d
+size 2334150
diff --git a/dev/tatoeba-dev.jpn-ukr.tsv b/dev/tatoeba-dev.jpn-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..75aa51b917676d75ded3cb22cd114849281ece8d
--- /dev/null
+++ b/dev/tatoeba-dev.jpn-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7443744e34f05cde578eb4bc360bbc642328356848d5d70400d86b7bd7fd0c6
+size 2025
diff --git a/dev/tatoeba-dev.jpn-vie.tsv b/dev/tatoeba-dev.jpn-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..89c34d1c4734a462ea740087189c751a3023ec0b
--- /dev/null
+++ b/dev/tatoeba-dev.jpn-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f2fe82a421633e9ab6ee7d0bb9f19b394689ded23acc7f411e86c01178f8611
+size 69734
diff --git a/dev/tatoeba-dev.jpn-zho.tsv b/dev/tatoeba-dev.jpn-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b876baef20674d722d7591fd6b56aaea79395808
--- /dev/null
+++ b/dev/tatoeba-dev.jpn-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d0f45ad7fe6662979a9201b5b67671239398f37da54ecbec2e2bbb0230f5460
+size 283939
diff --git a/dev/tatoeba-dev.jpn_Hani-cmn_Hans.tsv b/dev/tatoeba-dev.jpn_Hani-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..450b1d13c6032dfcf1f25f1703aad4f95df731f0
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hani-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1654df940e50cc95339ab2cd5226bfefd38f25a5c3e1e973f4cce7993e55371
+size 19663
diff --git a/dev/tatoeba-dev.jpn_Hani-cmn_Hant.tsv b/dev/tatoeba-dev.jpn_Hani-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..92b2ecdb9a6b5bf3bfdc48a862b326cb7dcf6506
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hani-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0673fb3d5624d63a9f55fd8533548523af15835ab98879a8c4b2a5ab87705756
+size 13502
diff --git a/dev/tatoeba-dev.jpn_Hani-ind.tsv b/dev/tatoeba-dev.jpn_Hani-ind.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a1485ec7c61d6a424ee3e3c99e00dba58aac2f72
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hani-ind.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c6ad1c6c40a3e4d91316781f5786cce923275f089bb9bbc7455166ce44b8d7d
+size 6174
diff --git a/dev/tatoeba-dev.jpn_Hani-nld.tsv b/dev/tatoeba-dev.jpn_Hani-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7d1c9b97837880d861165d61e8b75ad4a97cd032
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hani-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0aa4bc8357eb5e9098ee10834a5eb0135a9bdc3ff1a259ab4c24b33588300575
+size 9513
diff --git a/dev/tatoeba-dev.jpn_Hani-pol.tsv b/dev/tatoeba-dev.jpn_Hani-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..05a9605b3648afe5b0c61961a347a760878e1037
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hani-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b702da06f7bc51297d9bfece8c11398a20acbed68bc2e336a60d35084745452f
+size 151313
diff --git a/dev/tatoeba-dev.jpn_Hani-por.tsv b/dev/tatoeba-dev.jpn_Hani-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a305b2a1c0b05678853c2166b7a16492b1383390
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hani-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a7059e4a1dba7edc36db289069072892649d670b7a8f76939481e31068eafb6
+size 14546
diff --git a/dev/tatoeba-dev.jpn_Hani-rus.tsv b/dev/tatoeba-dev.jpn_Hani-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e4359537b0fe09b7647527b823a3891994d0c2fb
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hani-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86a629e99572c0d6ea4989d150471b13403f8257a10ea9ac10f9683f9a32d2a8
+size 162924
diff --git a/dev/tatoeba-dev.jpn_Hani-spa.tsv b/dev/tatoeba-dev.jpn_Hani-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7e52f85560357f063f3be091bd2dc408172e2262
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hani-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6804fa1541812697a9d2535f610b0923bc47d26c338b156f668ba1ce8a0ff606
+size 226500
diff --git a/dev/tatoeba-dev.jpn_Hani-vie.tsv b/dev/tatoeba-dev.jpn_Hani-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9fa41f1ee71d7ea94cdfd1488fe154f78ba8165f
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hani-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36edf22fe2514a05ae61215c80042b13e7d912bbbcb64b86eecfbf9bd0532de6
+size 9241
diff --git a/dev/tatoeba-dev.jpn_Hani-yue_Hant.tsv b/dev/tatoeba-dev.jpn_Hani-yue_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..12ebd9cbcfdabced2b85a20eff9ebb140740e5fc
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hani-yue_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15d0d9d7a743f54c69dc089c3a52641c9543bb21bde6e46c51a6c25b1f87caf1
+size 1694
diff --git a/dev/tatoeba-dev.jpn_Hira-cmn_Hans.tsv b/dev/tatoeba-dev.jpn_Hira-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..98f6fd50d7b86376758e9a28d7eaa195156db6d3
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99fdc733941402a6edb44b89a106ba2390eec52b9a2c02efd01234a0c9028570
+size 148800
diff --git a/dev/tatoeba-dev.jpn_Hira-cmn_Hant.tsv b/dev/tatoeba-dev.jpn_Hira-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5498fd071906b1e7464f55f8710d95479629d7a3
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10df6777d6abf7af405a7f920131225b7ef1e67a366bde696ce115c8588d860e
+size 75365
diff --git a/dev/tatoeba-dev.jpn_Hira-ind.tsv b/dev/tatoeba-dev.jpn_Hira-ind.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..beab5b15dd4b12a76079080d04ab624a11fec9d6
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-ind.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fac3542dbe7f6e4fd3ba910762289b0254af48f276c71d70ae508780786da04b
+size 80855
diff --git a/dev/tatoeba-dev.jpn_Hira-nds.tsv b/dev/tatoeba-dev.jpn_Hira-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b45f75c5a9c94b2c2a835c6d3a514bc1d2ba79cd
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf478883dcb03590be67fa08dbba5a6bc468ba1e868b0c84a5a521ae4af73d8e
+size 4705
diff --git a/dev/tatoeba-dev.jpn_Hira-nld.tsv b/dev/tatoeba-dev.jpn_Hira-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..daff981e45e74138d910a351db24aeb31c021281
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28e39a38a2481d8a5ccd6228076f1936c859d26296019f80f86c9568fe3a21ed
+size 82569
diff --git a/dev/tatoeba-dev.jpn_Hira-pol.tsv b/dev/tatoeba-dev.jpn_Hira-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..de2dc8163d7fb47c8201304b0237f4eb3eb7c59d
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44840fecf46434222e54f0b8593aea645114f33aff44ffb130bc0a1d8a492675
+size 1378699
diff --git a/dev/tatoeba-dev.jpn_Hira-por.tsv b/dev/tatoeba-dev.jpn_Hira-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a8df7241b8406c568ee034f01aae3497e6fde401
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f893df44e90f4efd056adeaf076ff2bfee801516aa96068cf47827964eec419
+size 93140
diff --git a/dev/tatoeba-dev.jpn_Hira-rus.tsv b/dev/tatoeba-dev.jpn_Hira-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8db52681c1812fb3dcf5597586b8035ab2f70f41
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f10142df3d809f8c779f831d59b21ea487e88b82a96a54eccd175250fb94e24
+size 1716408
diff --git a/dev/tatoeba-dev.jpn_Hira-spa.tsv b/dev/tatoeba-dev.jpn_Hira-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c68a1fc3dc101fd7d95f62f97cffe2c7abda8014
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2d8db48c5ff9969b748a81c5b939ac2c322e20b8874815eff224aeeaf279a44
+size 2061563
diff --git a/dev/tatoeba-dev.jpn_Hira-vie.tsv b/dev/tatoeba-dev.jpn_Hira-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0b84cc42addd4b59b7c082ab6f06ae23654f52f4
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:249278940b8541672ca9e0816c835db60defc59b69ca1a5470a7f5f6d1789184
+size 59266
diff --git a/dev/tatoeba-dev.jpn_Hira-yue_Hans.tsv b/dev/tatoeba-dev.jpn_Hira-yue_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7e6b5c1f5eba7d5ceaadfc97a8c45e5df50c04ae
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-yue_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae7c6d767605c0b956bd5e96f2080a36b030c462a2b239253b73fa6550f7ce6a
+size 7979
diff --git a/dev/tatoeba-dev.jpn_Hira-yue_Hant.tsv b/dev/tatoeba-dev.jpn_Hira-yue_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fd8e2e72d835236cc00a040ff01050423921318b
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-yue_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78da072bbc1046ff2e8bfaf0fad25a8dd05c3ccf0cf3eda65d9f02d1cf3917ea
+size 5186
diff --git a/dev/tatoeba-dev.jpn_Hira-zsm_Latn.tsv b/dev/tatoeba-dev.jpn_Hira-zsm_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d5b7c9d053c3ecf0998fa4d6dc2b72bf831497d9
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Hira-zsm_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1568d37a9cbaed84d3cc0ebb0c8cc7d0f8037d4dafa3309eb79dd4f5891e43e7
+size 6612
diff --git a/dev/tatoeba-dev.jpn_Kana-cmn_Hans.tsv b/dev/tatoeba-dev.jpn_Kana-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c0551edd3fd9e1d81ff596f503549838d3cc5909
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Kana-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66d6e0c81259656adc13a34572b35f87fee16424bd91670c293bcb31c6797265
+size 5522
diff --git a/dev/tatoeba-dev.jpn_Kana-ind.tsv b/dev/tatoeba-dev.jpn_Kana-ind.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..09cbc212b83a2793b3b77b5f4bd5f9d488202617
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Kana-ind.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:236044e66ab97364448bd10d938412b962e330fdf0780b79ddd4adca57c5b743
+size 2451
diff --git a/dev/tatoeba-dev.jpn_Kana-nld.tsv b/dev/tatoeba-dev.jpn_Kana-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f10f125635d7202b238d5877421caa3d82d48a1d
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Kana-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42e47f622387e8ad0e95562b8dbc1f75ab676e3ebaed191e5fdfc90818e7e489
+size 3635
diff --git a/dev/tatoeba-dev.jpn_Kana-pol.tsv b/dev/tatoeba-dev.jpn_Kana-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f366476dad303ed13fbd221f8468853f3e62b77c
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Kana-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3de65e6bc721ad988cbbbfa0d64ca5aad14abf49d5837e2565d19bdb4e86d290
+size 22958
diff --git a/dev/tatoeba-dev.jpn_Kana-por.tsv b/dev/tatoeba-dev.jpn_Kana-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9403a455f24002437aa8ed6b1553dba8f1cb91bb
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Kana-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa0422ca1334538efcfb6d7c2e8e8027a5e342f7b7175889e84e47c9dcc29822
+size 5868
diff --git a/dev/tatoeba-dev.jpn_Kana-rus.tsv b/dev/tatoeba-dev.jpn_Kana-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e9161dd5154cc90d3af311e59f1d879f32d36fbc
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Kana-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ddaf2ebd86b4c6fcc170dcbe3eca9319238acb3dfbf3d73c5e509a690f349208
+size 64911
diff --git a/dev/tatoeba-dev.jpn_Kana-spa.tsv b/dev/tatoeba-dev.jpn_Kana-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..08d16aa588105a78e3c2f244eecbde2fd605f7da
--- /dev/null
+++ b/dev/tatoeba-dev.jpn_Kana-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87e8c0921910e68312c7c5e4f583e7e8f64adb47a6672347b9616c6b92ca3855
+size 45288
diff --git a/dev/tatoeba-dev.kab-kab.tsv b/dev/tatoeba-dev.kab-kab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dbf5b4f11da6d5e94477a3a701ea282832206a0b
--- /dev/null
+++ b/dev/tatoeba-dev.kab-kab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0e41d9ca32f0e031d8f016643b2212d2433f235b2d415441cbe88fa6dd9e81c
+size 24446
diff --git a/dev/tatoeba-dev.kaz-rus.tsv b/dev/tatoeba-dev.kaz-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f2547a41269043d130ed3f0853aa9db9a4cc6521
--- /dev/null
+++ b/dev/tatoeba-dev.kaz-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebb09b930ded568a4fad94cd1c5b86f7b5043024ef6ecf2363ba892a4c2e2aa8
+size 241995
diff --git a/dev/tatoeba-dev.lad-lat.tsv b/dev/tatoeba-dev.lad-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..30ca6b7c3bb2bee2613f67fe7f1803dc563907c5
--- /dev/null
+++ b/dev/tatoeba-dev.lad-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c553f9738af4c65df31d75786af0791624d9309144336502ffe45348f8ac040a
+size 2277
diff --git a/dev/tatoeba-dev.lad-lfn.tsv b/dev/tatoeba-dev.lad-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..632a3ac635d2c8476ab2772567e10928ff8ff458
--- /dev/null
+++ b/dev/tatoeba-dev.lad-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ed426e61b34e8f84a81f250e832d7a39780ffb53b141c8d1e3fc4e4efa023cc
+size 5230
diff --git a/dev/tatoeba-dev.lad-spa.tsv b/dev/tatoeba-dev.lad-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..868da03df38e9372308fca11a12de3c2a6316f79
--- /dev/null
+++ b/dev/tatoeba-dev.lad-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa068a9d0a0f2d5475f571b98661dd50f5da10b052186434b6a4a1b3e5e424b3
+size 1938
diff --git a/dev/tatoeba-dev.lad-yid.tsv b/dev/tatoeba-dev.lad-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9fe776f2de548f7c6b55c47c15a78daa49142e79
--- /dev/null
+++ b/dev/tatoeba-dev.lad-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2d7253e882255baad62f23f74158c06556705a43891da22c379c8ad2d41aad9
+size 8610
diff --git a/dev/tatoeba-dev.lad_Latn-lat_Latn.tsv b/dev/tatoeba-dev.lad_Latn-lat_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e9f6993568b8fb3fdcdfedcb8b91fd99b38eac9a
--- /dev/null
+++ b/dev/tatoeba-dev.lad_Latn-lat_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ec73938b739ac85e433ded03ab4ba02a20bed7eb4900dcd17c9cb82cf542e96
+size 1318
diff --git a/dev/tatoeba-dev.lad_Latn-lfn_Latn.tsv b/dev/tatoeba-dev.lad_Latn-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a600bb42e15f0ff8e9857d415626edeb2522ce1e
--- /dev/null
+++ b/dev/tatoeba-dev.lad_Latn-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cb8a000d2d864b40024a728513dbad50c10d18a0b65d9b67852332f37eb8ade
+size 1921
diff --git a/dev/tatoeba-dev.lad_Latn-yid.tsv b/dev/tatoeba-dev.lad_Latn-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1de1393998aa7509944e55cae8041a43d5d47ac8
--- /dev/null
+++ b/dev/tatoeba-dev.lad_Latn-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07ff64dc39d238cf0ef1f0df5698e51a42072ebf358345be446a0d40adb59f0e
+size 4412
diff --git a/dev/tatoeba-dev.lat-lfn.tsv b/dev/tatoeba-dev.lat-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d4213b90660f524eba7848eb2d91a029fc691585
--- /dev/null
+++ b/dev/tatoeba-dev.lat-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bca46c9d2b80c053567797195211092d5a427d4cd6f42962522e88b4c6054ddf
+size 5123
diff --git a/dev/tatoeba-dev.lat-por.tsv b/dev/tatoeba-dev.lat-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8d67ce05fb5c2a5ea49a9a28f304a49f92e9c98b
--- /dev/null
+++ b/dev/tatoeba-dev.lat-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:464a1e86780a36bc622b046bda8bee92c002935ebe118e229bf6d671e3f5c1bc
+size 1040251
diff --git a/dev/tatoeba-dev.lat-rus.tsv b/dev/tatoeba-dev.lat-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c4b0cd6a781cc5d24bfc9b67a94aa717e05d73a4
--- /dev/null
+++ b/dev/tatoeba-dev.lat-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:759ad90f9b7ae8523b43dc6fcc88cfa0be950132a82b93188b33590365734e98
+size 95812
diff --git a/dev/tatoeba-dev.lat-spa.tsv b/dev/tatoeba-dev.lat-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..697e319f11f22a5e5bde68a3ebaf5b0da46bcd53
--- /dev/null
+++ b/dev/tatoeba-dev.lat-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fe8961a95ff27572702e16870f599013ce17bbf82c2b84264620cb1c7a49f60
+size 114753
diff --git a/dev/tatoeba-dev.lat-tlh.tsv b/dev/tatoeba-dev.lat-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c4ff6b3c55ad848dcf55e8ef272406a515fffb6b
--- /dev/null
+++ b/dev/tatoeba-dev.lat-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93f03583010c50c65d295a98dd6e9f9aa8411e511d1a8a93c8b46bcbe32326de
+size 1753
diff --git a/dev/tatoeba-dev.lat-yid.tsv b/dev/tatoeba-dev.lat-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b11ec4bd7583da270077b4e2d90ce7115c7054d7
--- /dev/null
+++ b/dev/tatoeba-dev.lat-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff031dcae1b040088394745d4f65088fd5959b7c0eb92d64e34d17e89ca15684
+size 6791
diff --git a/dev/tatoeba-dev.lat_Latn-lfn_Cyrl.tsv b/dev/tatoeba-dev.lat_Latn-lfn_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1a911989d50c130c53126db8f4bc0a776aa94a3f
--- /dev/null
+++ b/dev/tatoeba-dev.lat_Latn-lfn_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6849882de77b92d0315a0c2a9e0d6f935f5ae964594147c3f0fbfb536802f849
+size 1865
diff --git a/dev/tatoeba-dev.lat_Latn-lfn_Latn.tsv b/dev/tatoeba-dev.lat_Latn-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..781f81fe0619b721b5465133211172dc82fb87d7
--- /dev/null
+++ b/dev/tatoeba-dev.lat_Latn-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da6a20de61227f6d731370b40bdb7425f3ca16db920aac42c0433841f4e38c8d
+size 3258
diff --git a/dev/tatoeba-dev.lfn-por.tsv b/dev/tatoeba-dev.lfn-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..50e62bc822c11cdffba138f6d50dc276832a2608
--- /dev/null
+++ b/dev/tatoeba-dev.lfn-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a413983cc12470011b6785dc2df109460b3cc9943a8acc658a300b908f67a497
+size 135279
diff --git a/dev/tatoeba-dev.lfn-rus.tsv b/dev/tatoeba-dev.lfn-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b14a229c7f92da942fe593467d81860a05f86fde
--- /dev/null
+++ b/dev/tatoeba-dev.lfn-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14f252c031d1b63d483e3be67eca9fb23609b5a4488fae85c5def7a6e9419d75
+size 3459
diff --git a/dev/tatoeba-dev.lfn-spa.tsv b/dev/tatoeba-dev.lfn-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a0c469470e04c52e7d004a4ac1c9aa604da8cb42
--- /dev/null
+++ b/dev/tatoeba-dev.lfn-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cba240fd57f25af9c99be21ebd16766d2ec011deac65c242bdf485c0bde890dc
+size 5148
diff --git a/dev/tatoeba-dev.lfn-yid.tsv b/dev/tatoeba-dev.lfn-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8caec86ae064a28726f5bd7c1cd59accfb79b244
--- /dev/null
+++ b/dev/tatoeba-dev.lfn-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb913cf9756f7e331eeb72c712952f6fdb2a67f549fb397bc741ff1d40ae6a7c
+size 24927
diff --git a/dev/tatoeba-dev.lfn_Cyrl-por.tsv b/dev/tatoeba-dev.lfn_Cyrl-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e7b71715bd5c89518562df0d63ffb6ba37de5f30
--- /dev/null
+++ b/dev/tatoeba-dev.lfn_Cyrl-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a92dfa102ba2794c19c09083df585d5a5e778c618406b558eacc854de31b38c
+size 66291
diff --git a/dev/tatoeba-dev.lfn_Cyrl-spa.tsv b/dev/tatoeba-dev.lfn_Cyrl-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2b7027bf17a485e6ebd1ce354868ca79adeed273
--- /dev/null
+++ b/dev/tatoeba-dev.lfn_Cyrl-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3fa10cf460f61eb54c8baf9077d333b37a3026936a9d1eb2347003785a7a9aa
+size 2638
diff --git a/dev/tatoeba-dev.lfn_Cyrl-yid.tsv b/dev/tatoeba-dev.lfn_Cyrl-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a5a1283cc764f41a99094a4da194a6cd05249753
--- /dev/null
+++ b/dev/tatoeba-dev.lfn_Cyrl-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dac1bb714f7841c2a8490bc187bc54dd3a5a9737290acda4a491ddbab20dd83a
+size 6829
diff --git a/dev/tatoeba-dev.lfn_Latn-por.tsv b/dev/tatoeba-dev.lfn_Latn-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..515a6624b123a23a2d54c710726306a3da05afd9
--- /dev/null
+++ b/dev/tatoeba-dev.lfn_Latn-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3f8125c782cb60a3f10d1ab1cc3ffccf51e48d4db67b2761b1dda095adbc269
+size 68988
diff --git a/dev/tatoeba-dev.lfn_Latn-spa.tsv b/dev/tatoeba-dev.lfn_Latn-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..af920f2e43753656e39df60de2f986f6ef29472b
--- /dev/null
+++ b/dev/tatoeba-dev.lfn_Latn-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e699799fc71c020f9e1f4c0cd9829144b7675aa4e5f7d0c1d8337834161f6084
+size 2510
diff --git a/dev/tatoeba-dev.lfn_Latn-yid.tsv b/dev/tatoeba-dev.lfn_Latn-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..17700019e804433a63e1e0abd4e4f48931d2dc7e
--- /dev/null
+++ b/dev/tatoeba-dev.lfn_Latn-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea02a35bc83f1b3949f758255885fc747aa85e08d23341fd4e7f20f1f6bda17e
+size 18098
diff --git a/dev/tatoeba-dev.lit-rus.tsv b/dev/tatoeba-dev.lit-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..86e89237a3c4f45d320225d8fd454385441fe760
--- /dev/null
+++ b/dev/tatoeba-dev.lit-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbc5c74d11db131602ea7253799c01fc746578ff65ecd83fd57ec68bb588d3c7
+size 482377
diff --git a/dev/tatoeba-dev.lzh_Hani-cmn_Hans.tsv b/dev/tatoeba-dev.lzh_Hani-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..caf54f4e1371f210101f4d48724784a816d78dc6
--- /dev/null
+++ b/dev/tatoeba-dev.lzh_Hani-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60afe5df4545baa3e4fb14d99dbddc4ba00cb3e2abe5201e2f35b3f0e39cef7b
+size 2221
diff --git a/dev/tatoeba-dev.lzh_Hani-cmn_Hant.tsv b/dev/tatoeba-dev.lzh_Hani-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8f8ce8b83c1fa79b3d2a8c94d2a7698b1e74211d
--- /dev/null
+++ b/dev/tatoeba-dev.lzh_Hani-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abc2cf822c83c5dc0643ae5d92f76e2d51cad33066d34c76505fdf38ffc9cfea
+size 3871
diff --git a/dev/tatoeba-dev.nds-nld.tsv b/dev/tatoeba-dev.nds-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..41751cfbed6ac740ac72ff4a4fc44af94bc5dd1b
--- /dev/null
+++ b/dev/tatoeba-dev.nds-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21655d529bf8b00474189cc4dabc46c1a4df1ef8d4255781ec0093237ddd661a
+size 70369
diff --git a/dev/tatoeba-dev.nds-rus.tsv b/dev/tatoeba-dev.nds-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5af7dfb17f74f23347a67dd8f31397a9aba8295d
--- /dev/null
+++ b/dev/tatoeba-dev.nds-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7352821912e51f63246fa2916ed72a560a9aaa3a0f86ade621ec7dc00be99d1
+size 3359
diff --git a/dev/tatoeba-dev.nld-cmn_Hans.tsv b/dev/tatoeba-dev.nld-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..09a71af2542f508eaec02dd98d0926102916de07
--- /dev/null
+++ b/dev/tatoeba-dev.nld-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2bda0ec5d4bf6834e916086841e463ed9a4946f0c9a57e3cc64d9e99e82b34c
+size 3049
diff --git a/dev/tatoeba-dev.nld-nld.tsv b/dev/tatoeba-dev.nld-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e24ff352dbad82ba2a8ff972aa604f7d89d192ae
--- /dev/null
+++ b/dev/tatoeba-dev.nld-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:522b0b000a12fd1f4f56218afa4299a4626eafe5e12e9b37a628ea3532212403
+size 73790
diff --git a/dev/tatoeba-dev.nld-pol.tsv b/dev/tatoeba-dev.nld-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4349f2d5f0e247dc14ef2b91bf0cc9e2e9ada76d
--- /dev/null
+++ b/dev/tatoeba-dev.nld-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f6f98866df00f607c415a194c66750d33d60e55a34ba3b56e062265a8aa8e46
+size 73045
diff --git a/dev/tatoeba-dev.nld-por.tsv b/dev/tatoeba-dev.nld-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9d4d5bb8daebba8349ad65a203103f8d75714e7d
--- /dev/null
+++ b/dev/tatoeba-dev.nld-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67af69eaea82c939532bd2b83c32c8813e3c05e22941a4d754c527c705814fde
+size 365339
diff --git a/dev/tatoeba-dev.nld-ron.tsv b/dev/tatoeba-dev.nld-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..733a3488c0c13805990bf2bf21fb34a458c35c7f
--- /dev/null
+++ b/dev/tatoeba-dev.nld-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f363610ec8bf9c8dbb673e6203f443eae1e4385025eef924db895cc5a5ad650e
+size 77487
diff --git a/dev/tatoeba-dev.nld-rus.tsv b/dev/tatoeba-dev.nld-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ecb01aad268106e30080ad05fdea0d6a2646c30e
--- /dev/null
+++ b/dev/tatoeba-dev.nld-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4336cae40c0908b45ca25d094b53fafe2d491406d370d248918659f83c19fce2
+size 613374
diff --git a/dev/tatoeba-dev.nld-spa.tsv b/dev/tatoeba-dev.nld-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b72d683d55f32b3e640ffe03e82adbaae44641d4
--- /dev/null
+++ b/dev/tatoeba-dev.nld-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:830a98d2b07ba23da5a72160a33ade1d19ecfdb4ca0370b31268d7e3dfd096d5
+size 1450714
diff --git a/dev/tatoeba-dev.nld-tur.tsv b/dev/tatoeba-dev.nld-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c8bd29148798c1c40e82c078a20fb44e4b819f30
--- /dev/null
+++ b/dev/tatoeba-dev.nld-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b02695c822de54d531c62d5c809af131fe07cf2c0ad6b8c54edbeca7157d047
+size 286529
diff --git a/dev/tatoeba-dev.nld-ukr.tsv b/dev/tatoeba-dev.nld-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7b702ffed22b4ff2bc16b5850f3c6579b90bd6f0
--- /dev/null
+++ b/dev/tatoeba-dev.nld-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8deb71bfdb02a9f035ea8f80d54185299f89bdb551342b83b33c78f1f6780f44
+size 392653
diff --git a/dev/tatoeba-dev.nld-zho.tsv b/dev/tatoeba-dev.nld-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c961db27bcc35b175b7b3ece685100a4d0ce3f06
--- /dev/null
+++ b/dev/tatoeba-dev.nld-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f813838a39d325df92be851d52944d435d0d910570bac7aaad6b27fef75b5b96
+size 3797
diff --git a/dev/tatoeba-dev.nob-nob.tsv b/dev/tatoeba-dev.nob-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e580ccae0eac625f72dd1e5b5b9c04a544f76aed
--- /dev/null
+++ b/dev/tatoeba-dev.nob-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50f87cee4c5d07ac09e9879e7bb560807118bd95777a000c8c8e8c53641fb6b0
+size 17942
diff --git a/dev/tatoeba-dev.nor-nor.tsv b/dev/tatoeba-dev.nor-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0b50bd2ebecff0f0b9e313df4cefcecd057b297d
--- /dev/null
+++ b/dev/tatoeba-dev.nor-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a39a63a5baa928a5bbe2cc1173eca664b26096fd26fd87f0d83b9f4d2b6791b
+size 21110
diff --git a/dev/tatoeba-dev.ota-tur.tsv b/dev/tatoeba-dev.ota-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3b4405743a1d979d117eff0342fb95f80247c169
--- /dev/null
+++ b/dev/tatoeba-dev.ota-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fec141ece8b68d9c68c935829e31b8bd4f24307e36c1e158721e36defa2c723b
+size 2304
diff --git a/dev/tatoeba-dev.pol-rus.tsv b/dev/tatoeba-dev.pol-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d844d38b8ae30bab2cc4926d831b5a64f9a56e09
--- /dev/null
+++ b/dev/tatoeba-dev.pol-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ada19c3163fc34ae62b119d18931c9100bb5f20d433a81c3b54f4c4dfcf6ced
+size 96002
diff --git a/dev/tatoeba-dev.pol-spa.tsv b/dev/tatoeba-dev.pol-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ae8b0e48c453315f2ac6a0b690d500f3271a159e
--- /dev/null
+++ b/dev/tatoeba-dev.pol-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b4b84febc66fd26fc9163617f4e3bc8821036a6def4006d75db715ba7c142a2
+size 394091
diff --git a/dev/tatoeba-dev.pol-ukr.tsv b/dev/tatoeba-dev.pol-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..461e11c5d34ba3bd09418628afc29afb84e9cc65
--- /dev/null
+++ b/dev/tatoeba-dev.pol-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9bd6749e9c07b8258ebf182fbdb00df0ae8773f3e7b933ff0275d115048b5b5
+size 550111
diff --git a/dev/tatoeba-dev.por-por.tsv b/dev/tatoeba-dev.por-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e57064c586fa7f696a99cdd30740f3032309e665
--- /dev/null
+++ b/dev/tatoeba-dev.por-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:edc63518fd61683b7f6a99f7de73668fe0701165948a83054415ee3fbfb623df
+size 376940
diff --git a/dev/tatoeba-dev.por-ron.tsv b/dev/tatoeba-dev.por-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f3b4be4c220448a76f101c56776b18aae0f9858f
--- /dev/null
+++ b/dev/tatoeba-dev.por-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e14c472c33b81f6b518e3c123766e750dfd940db4d6119c726f9e9d3a2ea2da
+size 1971
diff --git a/dev/tatoeba-dev.por-rus.tsv b/dev/tatoeba-dev.por-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..23331aa908c247bd80b684b3ddb3b7b51850b35e
--- /dev/null
+++ b/dev/tatoeba-dev.por-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7325346ba264fc7f5dd7dc8b847c8c827ebe0eeaec1f88c840bebb01114a9f01
+size 1048814
diff --git a/dev/tatoeba-dev.por-spa.tsv b/dev/tatoeba-dev.por-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bb8bed46da5aeabb6f64696e8b1a18a053f95750
--- /dev/null
+++ b/dev/tatoeba-dev.por-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab709382477c990dcd04a1b6f8740bd2a445510302ddd53d19c3e53249ad1ce9
+size 4774470
diff --git a/dev/tatoeba-dev.por-toki.tsv b/dev/tatoeba-dev.por-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fffe6f9cbb4f2c52b32a234a67a928b6afb7ef7b
--- /dev/null
+++ b/dev/tatoeba-dev.por-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1118d4abe5f2c31efa2cc12f89b118d7134c2843c79010f2c68a4572a577d75d
+size 58547
diff --git a/dev/tatoeba-dev.por-ukr.tsv b/dev/tatoeba-dev.por-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..066d4e52a854be7d8436dbfe16bddabc028a98e6
--- /dev/null
+++ b/dev/tatoeba-dev.por-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5ed9bba428e1f81c3ff87286855f2db9d675d62e0e0ec7af10429d006800f66
+size 87842
diff --git a/dev/tatoeba-dev.ron-tur.tsv b/dev/tatoeba-dev.ron-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b413fe74c95292ff4d2ca2cd500862d4669c0298
--- /dev/null
+++ b/dev/tatoeba-dev.ron-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92c3419587f58dce739fb67dd4a6e7332056359f00e1e0ac7b1b2b91c13a05e8
+size 82967
diff --git a/dev/tatoeba-dev.rus-cmn_Hans.tsv b/dev/tatoeba-dev.rus-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8617dcf2776c7a779685933ed07c56aa827faed9
--- /dev/null
+++ b/dev/tatoeba-dev.rus-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2514dc4638db052774c9c0dc3fd5147c2f288493de74342ae9647e86605bdcf
+size 239435
diff --git a/dev/tatoeba-dev.rus-cmn_Hant.tsv b/dev/tatoeba-dev.rus-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a0b7d10b5967710947ad0bd080136f3be24f1e71
--- /dev/null
+++ b/dev/tatoeba-dev.rus-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0add2f744acb67ae6bc1816a6e98aad4d7093fb6b197bf72d8de7116ea3fac6c
+size 143626
diff --git a/dev/tatoeba-dev.rus-lzh_Bopo.tsv b/dev/tatoeba-dev.rus-lzh_Bopo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..12bef528eb95c0808451df1bdb73dacfce43e297
--- /dev/null
+++ b/dev/tatoeba-dev.rus-lzh_Bopo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66d9db5c57a8912d1306e2cb77e884e4204e24a639ed045f49df9d32f1debb16
+size 2674
diff --git a/dev/tatoeba-dev.rus-lzh_Hang.tsv b/dev/tatoeba-dev.rus-lzh_Hang.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8a0dfabdd43f03302b65243464467694dc5f9253
--- /dev/null
+++ b/dev/tatoeba-dev.rus-lzh_Hang.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a71f98d8f1e7a415590ef11e02265b4e8c0925d89f5a061a110c6be56d4c9330
+size 2064
diff --git a/dev/tatoeba-dev.rus-lzh_Hani.tsv b/dev/tatoeba-dev.rus-lzh_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ed2c6226a0e2e54e3903e8cfd9adeb7c4d7ad073
--- /dev/null
+++ b/dev/tatoeba-dev.rus-lzh_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:529c787b6835d4e4cb9b4a26922b6c2451a566c2a6b57888ee5dd18ae52d5730
+size 48296
diff --git a/dev/tatoeba-dev.rus-lzh_Hira.tsv b/dev/tatoeba-dev.rus-lzh_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d2615c444c064ab2779372dd165b8d55058408ec
--- /dev/null
+++ b/dev/tatoeba-dev.rus-lzh_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8db54c8dd112107b50f68773028c7a50d570f31570418450f7496ddc771c2015
+size 2037
diff --git a/dev/tatoeba-dev.rus-lzh_Kana.tsv b/dev/tatoeba-dev.rus-lzh_Kana.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e4ce00a59264e9725ec6b5b036297019f43a745b
--- /dev/null
+++ b/dev/tatoeba-dev.rus-lzh_Kana.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45b000dec3f7b7956ee81d1a02d3613ca14c35966ab2b9548a15bf294001bd97
+size 1795
diff --git a/dev/tatoeba-dev.rus-rus.tsv b/dev/tatoeba-dev.rus-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..03320cdd2be23f010077cb25b041112ee8c19c2e
--- /dev/null
+++ b/dev/tatoeba-dev.rus-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa183e2c1094173edfcf6f6548c042f2c8f6567f1165d840ca0a39735b0aa248
+size 189294
diff --git a/dev/tatoeba-dev.rus-slv.tsv b/dev/tatoeba-dev.rus-slv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8cb280c1682bf272811723a05626fee795b1d8ef
--- /dev/null
+++ b/dev/tatoeba-dev.rus-slv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4416c0b4f7988e0fb5393a250dad27e78b89f586c3af995054f9022bea373bc0
+size 4076
diff --git a/dev/tatoeba-dev.rus-spa.tsv b/dev/tatoeba-dev.rus-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..baa80e873151f7406c4cd4d70bd4c6175d7ca321
--- /dev/null
+++ b/dev/tatoeba-dev.rus-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4221b932f2c621512bfaa77a2caf8d1770aa0f76a5f7d9b9caaa3952e8c3a384
+size 8750687
diff --git a/dev/tatoeba-dev.rus-swe.tsv b/dev/tatoeba-dev.rus-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..07190a1c87ed0d29eb45c2251b676baa81562e5b
--- /dev/null
+++ b/dev/tatoeba-dev.rus-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a307933b334f1cbbfa656867c9303b8874c01c2f9de99504e73d4b99cc3359c
+size 4722
diff --git a/dev/tatoeba-dev.rus-tat.tsv b/dev/tatoeba-dev.rus-tat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..96be2b1ec84dcb8f1422fbedbdd8f716f4cd8c14
--- /dev/null
+++ b/dev/tatoeba-dev.rus-tat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d1431e060a917c8b0272a87abed9f8372d423d7099751b4cf5319e1c8212c77
+size 203708
diff --git a/dev/tatoeba-dev.rus-toki.tsv b/dev/tatoeba-dev.rus-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6c5a1d0c80df6327e8895b8094c7d2d64e918ad0
--- /dev/null
+++ b/dev/tatoeba-dev.rus-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e76292bb66b34402e6dde7a8e685c1187c47312aba3602c19ede89ce40cc4c50
+size 2708
diff --git a/dev/tatoeba-dev.rus-tur.tsv b/dev/tatoeba-dev.rus-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..70c7d7c6b3c9669a6ad396da59072c40f897434a
--- /dev/null
+++ b/dev/tatoeba-dev.rus-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e25ce626e7b146142d5ddc6bc6fcdb93145e5fcb09241954a9f57c4e3de4233f
+size 658089
diff --git a/dev/tatoeba-dev.rus-ukr.tsv b/dev/tatoeba-dev.rus-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6d989440375fc525b1e96f1a1386b5e46d08b254
--- /dev/null
+++ b/dev/tatoeba-dev.rus-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc9eee736214c8b97684ba7fefb3e94d2750085bac6e4c1f02fcd3b1765bc628
+size 831738
diff --git a/dev/tatoeba-dev.rus-yue_Hans.tsv b/dev/tatoeba-dev.rus-yue_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..825d56e6940db03fd7149d6de956bae34c9dd49e
--- /dev/null
+++ b/dev/tatoeba-dev.rus-yue_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10dc899cd7a11e1a530d7bf308534e29fdd0fdced1056cfccb542d08481249f4
+size 31847
diff --git a/dev/tatoeba-dev.rus-yue_Hant.tsv b/dev/tatoeba-dev.rus-yue_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e1187161c7175ec5d373846ed82009081ab36dcb
--- /dev/null
+++ b/dev/tatoeba-dev.rus-yue_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f981b7b1ea5742ed0e8e154685b2475de934f669311e7dcf7dae031647e1e89
+size 27263
diff --git a/dev/tatoeba-dev.rus-zho.tsv b/dev/tatoeba-dev.rus-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9f329d5d5359f76edc58f5ff78b72fd6545381ca
--- /dev/null
+++ b/dev/tatoeba-dev.rus-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:413d4e0da811106a26ed1bb3cc4c911017fd6e46f88d5ebf8a5e9cdb96a1b75f
+size 504643
diff --git a/dev/tatoeba-dev.slv-ukr.tsv b/dev/tatoeba-dev.slv-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..28e1afedbd3ab715d28a7c8635cf993a672a8cc6
--- /dev/null
+++ b/dev/tatoeba-dev.slv-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ef6d81230310019fdb93ddcc9c9a9db88db2c9011b6b5eed100929e7fb8a9bc
+size 3873
diff --git a/dev/tatoeba-dev.spa-cmn_Hans.tsv b/dev/tatoeba-dev.spa-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1e5fd167ed937f4cbf75330cf11b28fb28a04a59
--- /dev/null
+++ b/dev/tatoeba-dev.spa-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae7bb7296bfe9d6111bc4eefd014a80bfa58f2b7a1ba373e490e8a4dbd8ef452
+size 212686
diff --git a/dev/tatoeba-dev.spa-cmn_Hant.tsv b/dev/tatoeba-dev.spa-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b91de68c83ad68f467dc5a4c5930b664a2978696
--- /dev/null
+++ b/dev/tatoeba-dev.spa-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6774cb58f1e351acfaa45eb1df541b5767444c6a915b22cea18f444b77fa080b
+size 278564
diff --git a/dev/tatoeba-dev.spa-lzh_Hani.tsv b/dev/tatoeba-dev.spa-lzh_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d671115538bce25c23d3085ae39fe1e2e57ce836
--- /dev/null
+++ b/dev/tatoeba-dev.spa-lzh_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31aa245c3f877f600a9fded6f80f429bd9eaf4caff27f4fd826e7a50ed68bf17
+size 2005
diff --git a/dev/tatoeba-dev.spa-spa.tsv b/dev/tatoeba-dev.spa-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b65b0ab93e775d939ad94564cfe1c5f94072ee75
--- /dev/null
+++ b/dev/tatoeba-dev.spa-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8d2b0d59862d176ef27ca9768e01b77e8a9b4178ca54303904423f707182b78
+size 239817
diff --git a/dev/tatoeba-dev.spa-swe.tsv b/dev/tatoeba-dev.spa-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8aa0968d4c6ea9fa330646ee56e58c6a2288104a
--- /dev/null
+++ b/dev/tatoeba-dev.spa-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:099f08db521bf4950b52d5ab98e55ec93f3998dbcdc724ee435730fee31bb3a7
+size 14250
diff --git a/dev/tatoeba-dev.spa-toki.tsv b/dev/tatoeba-dev.spa-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..573bbcf118c370c6ab6c4bd2af820710b2f6f503
--- /dev/null
+++ b/dev/tatoeba-dev.spa-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfa0f56387e1daf86ff3831bc6df01c977ed69697385956389caf5c39fdc474a
+size 3933
diff --git a/dev/tatoeba-dev.spa-tur.tsv b/dev/tatoeba-dev.spa-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e1e9180d6dff4e4f3180f8c521ab6da31bc0cb3e
--- /dev/null
+++ b/dev/tatoeba-dev.spa-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cb805efba98dcb7300ed9e8bf13b9f09bf40425c8ded1accad4d3009d2733ea
+size 1263107
diff --git a/dev/tatoeba-dev.spa-ukr.tsv b/dev/tatoeba-dev.spa-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ad9cd41280544640ca90916f5d984f7e8e9651ff
--- /dev/null
+++ b/dev/tatoeba-dev.spa-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b917f7d58586e9e777b7f1abad7826a28380d09d00303f65d661ba8e00417f3c
+size 1051799
diff --git a/dev/tatoeba-dev.spa-wuu.tsv b/dev/tatoeba-dev.spa-wuu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b98a7bd34d2f96c9f412111bb4ec3b778f73333c
--- /dev/null
+++ b/dev/tatoeba-dev.spa-wuu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c4f48fb26896693469750a2d6de9f2804f9330278eeecbc9fa23aa379f260bb
+size 7795
diff --git a/dev/tatoeba-dev.spa-wuu_Hani.tsv b/dev/tatoeba-dev.spa-wuu_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b3c5ec96025a7bba284dc57603a4f6b90dc27ba3
--- /dev/null
+++ b/dev/tatoeba-dev.spa-wuu_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b153c109f9b04b56ee0d26c06a21ea63727d699b833c4000ad11c11fcbdf2b9a
+size 2776
diff --git a/dev/tatoeba-dev.spa-yid.tsv b/dev/tatoeba-dev.spa-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4e5216ffbd8d0633811e79caa97dd57c0755b6f8
--- /dev/null
+++ b/dev/tatoeba-dev.spa-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88bedcaa3970d17c335929abffbfe2440b0220a98ec8d3f81e3e132d5ef1e285
+size 4666
diff --git a/dev/tatoeba-dev.spa-yue_Hans.tsv b/dev/tatoeba-dev.spa-yue_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6935b2f5f73b8d1f1d3c3b178d4970d5074bec58
--- /dev/null
+++ b/dev/tatoeba-dev.spa-yue_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d93f099e35fa0a807af2ef4798332a86d8d247d3970a2c71f08ae0314264a95a
+size 2201
diff --git a/dev/tatoeba-dev.spa-yue_Hant.tsv b/dev/tatoeba-dev.spa-yue_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b50b4a0289d972f760f2800e57aa47e1603367b0
--- /dev/null
+++ b/dev/tatoeba-dev.spa-yue_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f58577e35a47d46afcc05f43f20a8128d3aed2144fd9cfab98436113aa4d90f0
+size 1482
diff --git a/dev/tatoeba-dev.spa-zho.tsv b/dev/tatoeba-dev.spa-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..28e5ea5d5d6ac77f41ceb6e75e86cf53d8ada947
--- /dev/null
+++ b/dev/tatoeba-dev.spa-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f5f9a0599afc552c1d31f6e2047f693928c896b91e2ca3af470d417e14e2ba3
+size 509628
diff --git a/dev/tatoeba-dev.srp_Cyrl-nob.tsv b/dev/tatoeba-dev.srp_Cyrl-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..63e4998e1dd8daf5805d2b709a4ecce5434aefba
--- /dev/null
+++ b/dev/tatoeba-dev.srp_Cyrl-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:920d01b4f84871ceefec18d40db5e615c50f8031610cb388eaf933cd50c3ae1d
+size 19905
diff --git a/dev/tatoeba-dev.srp_Cyrl-rus.tsv b/dev/tatoeba-dev.srp_Cyrl-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a1ce6a2a35d59d217e825855d966b17d611a63e1
--- /dev/null
+++ b/dev/tatoeba-dev.srp_Cyrl-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20ea46b72c96651fbd2eb60bc61e9977828dbe47993f3747602790cd749d771f
+size 166638
diff --git a/dev/tatoeba-dev.srp_Latn-nob.tsv b/dev/tatoeba-dev.srp_Latn-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e288b70668874157e0336e03d33c3e77a4c539cc
--- /dev/null
+++ b/dev/tatoeba-dev.srp_Latn-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b4944117a18c5af26aea012011f9433a19ca4e0e6af007a62cce0af0aa00aba
+size 550871
diff --git a/dev/tatoeba-dev.srp_Latn-rus.tsv b/dev/tatoeba-dev.srp_Latn-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b5ccf923455505d05a7e70cadb82664f14865295
--- /dev/null
+++ b/dev/tatoeba-dev.srp_Latn-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4cfc70164516f7534123c66f5740b0f5a416d3abf505a41f2b354201aca43ea7
+size 209425
diff --git a/dev/tatoeba-dev.tlh-yid.tsv b/dev/tatoeba-dev.tlh-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..974dac23364809e9e4095bd68c9fdec4c97ac5ee
--- /dev/null
+++ b/dev/tatoeba-dev.tlh-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f823f3c16f53bdb0c63693aa47b099481777f8dc70f7de8d9e359930f03a4d98
+size 4091
diff --git a/dev/tatoeba-dev.tur-tur.tsv b/dev/tatoeba-dev.tur-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..787f2ff06b120a01e1d87d4b26b8a1b6cfa42725
--- /dev/null
+++ b/dev/tatoeba-dev.tur-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15470f7beffcc5714e5d71d637b03ad389fea7f6db93c4f0a5018c71b37d69cc
+size 86899
diff --git a/dev/tatoeba-dev.tur-uig.tsv b/dev/tatoeba-dev.tur-uig.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8bd8b9318a9cc3bb4258309d97e3edc35e61ac15
--- /dev/null
+++ b/dev/tatoeba-dev.tur-uig.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d611fdc83d4e373dcecc848e3301a96c5a47989b6340bed0f4d18e9be223b5f0
+size 109693
diff --git a/dev/tatoeba-dev.tur-ukr.tsv b/dev/tatoeba-dev.tur-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3f08ef770f0cd86b052ebf1fb98c18d99f55cb82
--- /dev/null
+++ b/dev/tatoeba-dev.tur-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d63b35e178cb5217ccda1af6eb782081dfa4923da12c563467bed619ce784eb1
+size 510537
diff --git a/dev/tatoeba-dev.wuu-cmn_Hans.tsv b/dev/tatoeba-dev.wuu-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..86b5ae1a449e2958f3a186e63536bbda2cd0b93d
--- /dev/null
+++ b/dev/tatoeba-dev.wuu-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b6b900b4ad8e61771f5df9a17cac5ccfcc4b64ed963624fca8497d48120006c
+size 158748
diff --git a/dev/tatoeba-dev.wuu-cmn_Hant.tsv b/dev/tatoeba-dev.wuu-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..57c49380eee563d1c9372905ec031ab65e2253e3
--- /dev/null
+++ b/dev/tatoeba-dev.wuu-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:166abc1aa80e1c2a5bdf5aae3f2426538c0e091563ed37c2ae4ff715b4ed3f42
+size 30803
diff --git a/dev/tatoeba-dev.wuu-yue_Hans.tsv b/dev/tatoeba-dev.wuu-yue_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0472c7b46d3d1d2852e066ee7b771547298e0f9e
--- /dev/null
+++ b/dev/tatoeba-dev.wuu-yue_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9963edc2198b9accb86601eb6023def382a612515a10fe9e751c597ca14bb8b3
+size 2586
diff --git a/dev/tatoeba-dev.wuu-yue_Hant.tsv b/dev/tatoeba-dev.wuu-yue_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b060d2ac6f42c39241478f2353ea9b83f3b89011
--- /dev/null
+++ b/dev/tatoeba-dev.wuu-yue_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4d5a0f44ef0f30123a90a029b0b503683fea07e6f43ef7ad154e967d9575b59
+size 1582
diff --git a/dev/tatoeba-dev.wuu_Hani-cmn_Hans.tsv b/dev/tatoeba-dev.wuu_Hani-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c89447b01b6af6cb347bc7e234b424f66e28318f
--- /dev/null
+++ b/dev/tatoeba-dev.wuu_Hani-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8dac0bf6a55ef34b85d58ff41172db16eb0a9fe90b16a0a1cf9e6979ebe2f177
+size 4105
diff --git a/dev/tatoeba-dev.wuu_Hani-cmn_Hant.tsv b/dev/tatoeba-dev.wuu_Hani-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b6132b8a2356cfd0fc7289efbfb2e966eecbcb23
--- /dev/null
+++ b/dev/tatoeba-dev.wuu_Hani-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b771131ce7c23ad1c36c879b926b0eac6bdc1dbc1182942f1848dde6ac9d3938
+size 9189
diff --git a/dev/tatoeba-dev.yue_Hans-cmn_Hans.tsv b/dev/tatoeba-dev.yue_Hans-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..648d799995e382375768408a5a7a3726bebcfe5a
--- /dev/null
+++ b/dev/tatoeba-dev.yue_Hans-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be1f79e87c241262ae0541d50636a42baf8276b912b2bf4627c85b3350981110
+size 10889
diff --git a/dev/tatoeba-dev.yue_Hans-cmn_Hant.tsv b/dev/tatoeba-dev.yue_Hans-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8ec2f02e3ff3c4be5a5d9d177cecac2d3596b892
--- /dev/null
+++ b/dev/tatoeba-dev.yue_Hans-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b91c29549820f4d02e9eaddfe5995cc4b17c0bdfba26d073b8c05199a2e275b
+size 13361
diff --git a/dev/tatoeba-dev.yue_Hant-cmn_Hans.tsv b/dev/tatoeba-dev.yue_Hant-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7acbcbcd15fc09e6fe7013f431c7f6c1d7c3f151
--- /dev/null
+++ b/dev/tatoeba-dev.yue_Hant-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45aeab134fc410fc19009594a96358ea8d056e9f6fc7814723b9f0f4fa42b819
+size 5047
diff --git a/dev/tatoeba-dev.yue_Hant-cmn_Hant.tsv b/dev/tatoeba-dev.yue_Hant-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..afe0ebb778b21a409662d1280c849be1646f2d73
--- /dev/null
+++ b/dev/tatoeba-dev.yue_Hant-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:162b08c9d1d3941f621f5e136d909aa20d464af5c465c68bb5638ae33675df0f
+size 5947
diff --git a/dev/tatoeba-dev.zho-zho.tsv b/dev/tatoeba-dev.zho-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6a6fb04a39d2eca734593474446201cda46f5734
--- /dev/null
+++ b/dev/tatoeba-dev.zho-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9550af97e665f3c52ff956a1d031239cef119ba6f35cb33b10d2d169e8ac2a7
+size 502917
diff --git a/tatoeba_mt.py b/tatoeba_mt.py
new file mode 100644
index 0000000000000000000000000000000000000000..49dd0094f53ab90e0a5f18042f7ac2bbcfda570f
--- /dev/null
+++ b/tatoeba_mt.py
@@ -0,0 +1,140 @@
+# coding=utf-8
+# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""tatoeba-mt"""
+
+
+import os
+import csv
+import datasets
+
+
+_CITATION = """\
+@inproceedings{tiedemann-2020-tatoeba,
+    title = "The {T}atoeba {T}ranslation {C}hallenge {--} {R}ealistic Data Sets for Low Resource and Multilingual {MT}",
+    author = {Tiedemann, J{\"o}rg},
+    booktitle = "Proceedings of the Fifth Conference on Machine Translation",
+    month = nov,
+    year = "2020",
+    publisher = "Association for Computational Linguistics",
+    url = "https://aclanthology.org/2020.wmt-1.139",
+    pages = "1174--1182",
+}
+"""
+
+_DESCRIPTION = """\
+The Tatoeba Translation Challenge is a multilingual data set of
+machine translation benchmarks derived from user-contributed
+translations collected by [Tatoeba.org](https://tatoeba.org/) and
+provided as parallel corpus from [OPUS](https://opus.nlpl.eu/). This
+dataset includes test and development data sorted by language pair. It
+includes test sets for hundreds of language pairs and is continuously
+updated. Please, check the version number tag to refer to the release
+that your are using.
+"""
+
+_URL = {
+    "evaluation": "https://object.pouta.csc.fi/Tatoeba-Challenge-devtest/test.tar",
+    "development": "https://object.pouta.csc.fi/Tatoeba-Challenge-devtest/dev.tar",
+}
+
+_RELEASE = "v2021-08-07"
+_RELEASES = [ "v2020-07-28", "v2021-03-30", "v2021-08-07", ]
+_LanguagePairs = [ "afr-deu", "afr-eng", "afr-epo", "afr-nld", "afr-rus", "afr-spa", "ain-fin", "ara-ber", "ara-ber_Latn", "ara-deu", "ara-ell", "ara-eng", "ara-epo", "ara-fra", "ara-heb", "ara-ita", "ara-jpn", "ara-jpn_Hira", "ara-pol", "ara-rus", "ara-spa", "ara-tur", "arq-eng", "avk-fra", "avk-spa", "awa-eng", "aze-eng", "aze-spa", "aze-tur", "aze_Latn-tur", "bel-deu", "bel-eng", "bel-epo", "bel-fra", "bel-ita", "bel-lat", "bel-nld", "bel-pol", "bel-rus", "bel-spa", "bel-ukr", "bel-zho", "ben-eng", "ber-deu", "ber-eng", "ber-epo", "ber-fra", "ber-spa", "ber_Latn-deu", "ber_Latn-eng", "ber_Latn-epo", "ber_Latn-fra", "bre-eng", "bre-fra", "bua-rus", "bua_Cyrl-rus", "bul-bul", "bul-cmn_Hans", "bul-deu", "bul-eng", "bul-epo", "bul-fra", "bul-ita", "bul-jpn", "bul-jpn_Hira", "bul-rus", "bul-spa", "bul-tur", "bul-ukr", "bul-zho", "cat-deu", "cat-eng", "cat-epo", "cat-fra", "cat-ita", "cat-nld", "cat-por", "cat-spa", "cat-ukr", "cbk-eng", "ceb-deu", "ceb-eng", "ces-deu", "ces-eng", "ces-epo", "ces-fra", "ces-hun", "ces-ita", "ces-lat", "ces-pol", "ces-rus", "ces-slv", "ces-spa", "ces-ukr", "cha-eng", "chm-rus", "chv-eng", "chv-rus", "chv-tur", "cmn_Hans-wuu", "cor-deu", "cor-eng", "cor-epo", "cor-fra", "cor-ita", "cor-rus", "cor-spa", "crh-tur", "cym-eng", "dan-dan", "dan-deu", "dan-eng", "dan-epo", "dan-fin", "dan-fra", "dan-ita", "dan-jpn", "dan-jpn_Hira", "dan-nld", "dan-nob", "dan-nor", "dan-por", "dan-rus", "dan-spa", "dan-swe", "dan-tur", "deu-cmn_Hans", "deu-cmn_Hant", "deu-deu", "deu-dsb", "deu-ell", "deu-eng", "deu-epo", "deu-est", "deu-eus", "deu-fas", "deu-fin", "deu-fra", "deu-frr", "deu-gos", "deu-hbs", "deu-heb", "deu-hrv", "deu-hrx", "deu-hsb", "deu-hun", "deu-ido", "deu-ile", "deu-ina", "deu-ind", "deu-isl", "deu-ita", "deu-jbo", "deu-jpn", "deu-jpn_Hani", "deu-jpn_Hira", "deu-jpn_Kana", "deu-kab", "deu-kor", "deu-kor_Hang", "deu-kur", "deu-kur_Latn", "deu-lad", "deu-lat", "deu-lfn", "deu-lfn_Latn", "deu-lit", "deu-ltz", "deu-msa", "deu-nds", "deu-nld", "deu-nob", "deu-nor", "deu-pol", "deu-por", "deu-ron", "deu-run", "deu-rus", "deu-slv", "deu-spa", "deu-srp_Latn", "deu-swe", "deu-swg", "deu-tat", "deu-tgl", "deu-tlh", "deu-toki", "deu-tur", "deu-ukr", "deu-vie", "deu-vol", "deu-yid", "deu-zho", "dsb-hsb", "dsb-slv", "dtp-eng", "dtp-jpn", "dtp-jpn_Hira", "dtp-msa", "dtp-zsm_Latn", "egl-ita", "ell-ell", "ell-eng", "ell-epo", "ell-fra", "ell-ita", "ell-nld", "ell-por", "ell-rus", "ell-spa", "ell-swe", "ell-tur", "eng-bos_Latn", "eng-cmn_Hans", "eng-cmn_Hant", "eng-eng", "eng-epo", "eng-est", "eng-eus", "eng-fao", "eng-fas", "eng-fin", "eng-fra", "eng-fry", "eng-gla", "eng-gle", "eng-glg", "eng-gos", "eng-got", "eng-grc", "eng-gsw", "eng-hbs", "eng-heb", "eng-hin", "eng-hoc", "eng-hoc_Latn", "eng-hrv", "eng-hrx", "eng-hun", "eng-hye", "eng-ido", "eng-ido_Latn", "eng-ile", "eng-ilo", "eng-ina", "eng-ind", "eng-isl", "eng-ita", "eng-jav", "eng-jbo", "eng-jbo_Latn", "eng-jpn", "eng-jpn_Hani", "eng-jpn_Hira", "eng-jpn_Kana", "eng-kab", "eng-kat", "eng-kaz", "eng-kaz_Cyrl", "eng-kha", "eng-khm", "eng-kor", "eng-kor_Hang", "eng-kur", "eng-kur_Latn", "eng-kzj", "eng-lad", "eng-lad_Latn", "eng-lat", "eng-lav", "eng-lfn", "eng-lfn_Cyrl", "eng-lfn_Latn", "eng-lit", "eng-ltz", "eng-mal", "eng-mar", "eng-mkd", "eng-mlt", "eng-mon", "eng-mri", "eng-msa", "eng-mya", "eng-nds", "eng-nld", "eng-nno", "eng-nob", "eng-nor", "eng-nov", "eng-nst", "eng-oci", "eng-orv", "eng-ota", "eng-ota_Arab", "eng-ota_Latn", "eng-pam", "eng-pes", "eng-pms", "eng-pol", "eng-por", "eng-prg", "eng-que", "eng-rom", "eng-ron", "eng-run", "eng-rus", "eng-slv", "eng-spa", "eng-sqi", "eng-srp_Cyrl", "eng-srp_Latn", "eng-swa", "eng-swe", "eng-tam", "eng-tat", "eng-tel", "eng-tgl", "eng-tha", "eng-tlh", "eng-toki", "eng-tuk", "eng-tuk_Latn", "eng-tur", "eng-tzl", "eng-tzl_Latn", "eng-uig", "eng-uig_Arab", "eng-ukr", "eng-urd", "eng-uzb", "eng-uzb_Latn", "eng-vie", "eng-vol", "eng-war", "eng-xal", "eng-yid", "eng-yue_Hans", "eng-yue_Hant", "eng-zho", "eng-zsm_Latn", "eng-zza", "epo-cmn_Hans", "epo-cmn_Hant", "epo-epo", "epo-fas", "epo-fin", "epo-fra", "epo-glg", "epo-hbs", "epo-heb", "epo-hrv", "epo-hun", "epo-ido", "epo-ile", "epo-ile_Latn", "epo-ina", "epo-isl", "epo-ita", "epo-jbo", "epo-jpn", "epo-jpn_Hani", "epo-jpn_Hira", "epo-lad", "epo-lad_Latn", "epo-lat", "epo-lfn", "epo-lfn_Latn", "epo-lit", "epo-nds", "epo-nld", "epo-nob", "epo-nor", "epo-oci", "epo-pol", "epo-por", "epo-ron", "epo-rus", "epo-slv", "epo-spa", "epo-srp_Cyrl", "epo-srp_Latn", "epo-swe", "epo-tgl", "epo-tlh", "epo-toki", "epo-tur", "epo-ukr", "epo-vie", "epo-vol", "epo-yid", "epo-zho", "est-rus", "eus-jpn", "eus-rus", "eus-spa", "fas-fra", "fin-fin", "fin-fkv", "fin-fra", "fin-heb", "fin-hun", "fin-ita", "fin-jpn", "fin-jpn_Hani", "fin-jpn_Hira", "fin-jpn_Kana", "fin-kor", "fin-kor_Hang", "fin-kur", "fin-lat", "fin-nld", "fin-nno", "fin-nob", "fin-nor", "fin-pol", "fin-por", "fin-rus", "fin-spa", "fin-swe", "fin-tur", "fin-zho", "fra-cmn_Hans", "fra-cmn_Hant", "fra-fra", "fra-gcf", "fra-hbs", "fra-heb", "fra-hrv", "fra-hun", "fra-ido", "fra-ile", "fra-ina", "fra-ind", "fra-ita", "fra-jbo", "fra-jpn", "fra-jpn_Hani", "fra-jpn_Hira", "fra-kab", "fra-kor", "fra-kor_Hang", "fra-lat", "fra-lfn", "fra-lfn_Latn", "fra-msa", "fra-nds", "fra-nld", "fra-nob", "fra-nor", "fra-oci", "fra-pcd", "fra-pol", "fra-por", "fra-ron", "fra-run", "fra-rus", "fra-slv", "fra-spa", "fra-swe", "fra-tat", "fra-tgl", "fra-tlh", "fra-tlh_Latn", "fra-toki", "fra-toki_Latn", "fra-tur", "fra-uig", "fra-uig_Arab", "fra-ukr", "fra-vie", "fra-wuu", "fra-yid", "fra-zho", "fry-nld", "gcf-gcf", "gla-spa", "glg-por", "glg-spa", "gos-nld", "grn-por", "grn-spa", "hbs-ita", "hbs-jpn", "hbs-nor", "hbs-pol", "hbs-rus", "hbs-spa", "hbs-ukr", "hbs-zho", "heb-cmn_Hans", "heb-cmn_Hant", "heb-heb", "heb-hun", "heb-ina", "heb-ita", "heb-jpn", "heb-jpn_Hira", "heb-lad", "heb-lat", "heb-lfn", "heb-lfn_Latn", "heb-nld", "heb-pol", "heb-por", "heb-rus", "heb-spa", "heb-tur", "heb-ukr", "heb-yid", "heb-zho", "hin-urd", "hin-zho", "hrv-jpn_Hira", "hrv-pol", "hrv-spa", "hrv-ukr", "hsb-slv", "hun-cmn_Hans", "hun-hun", "hun-ita", "hun-jpn", "hun-jpn_Hani", "hun-jpn_Hira", "hun-kor", "hun-kor_Hang", "hun-lat", "hun-nld", "hun-pol", "hun-por", "hun-rus", "hun-spa", "hun-swe", "hun-tur", "hun-ukr", "hun-zho", "hye-rus", "ido-ina", "ido-ita", "ido-lfn", "ido-spa", "ido-yid", "ido_Latn-lfn_Latn", "ina-ita", "ina-lad", "ina-lat", "ina-lfn", "ina-nld", "ina-por", "ina-rus", "ina-spa", "ina-tlh", "ina-tur", "ina-yid", "ina_Latn-lad_Latn", "ina_Latn-lfn_Latn", "ina_Latn-tlh_Latn", "ind-zsm_Latn", "isl-ita", "isl-jpn", "isl-jpn_Hira", "isl-spa", "ita-cmn_Hans", "ita-cmn_Hant", "ita-ind", "ita-ita", "ita-jpn", "ita-jpn_Hani", "ita-jpn_Hira", "ita-lat", "ita-lit", "ita-msa", "ita-nds", "ita-nld", "ita-nor", "ita-pms", "ita-pol", "ita-por", "ita-ron", "ita-rus", "ita-spa", "ita-swe", "ita-toki", "ita-tur", "ita-ukr", "ita-vie", "ita-yid", "ita-zho", "jbo-jpn", "jbo-rus", "jbo-spa", "jbo-swe", "jbo-zho", "jbo_Latn-cmn_Hans", "jbo_Latn-cmn_Hant", "jbo_Latn-jpn_Hira", "jpn-jpn", "jpn-kor", "jpn-lit", "jpn-mar", "jpn-msa", "jpn-nds", "jpn-nld", "jpn-nor", "jpn-pol", "jpn-por", "jpn-rus", "jpn-spa", "jpn-swe", "jpn-tlh", "jpn-toki", "jpn-tur", "jpn-ukr", "jpn-vie", "jpn-zho", "jpn_Hani-cmn_Hans", "jpn_Hani-nld", "jpn_Hani-pol", "jpn_Hani-por", "jpn_Hani-rus", "jpn_Hani-spa", "jpn_Hira-cmn_Hans", "jpn_Hira-cmn_Hant", "jpn_Hira-ind", "jpn_Hira-jpn_Hira", "jpn_Hira-kor_Hang", "jpn_Hira-lit", "jpn_Hira-mar", "jpn_Hira-nds", "jpn_Hira-nld", "jpn_Hira-nob", "jpn_Hira-pol", "jpn_Hira-por", "jpn_Hira-rus", "jpn_Hira-spa", "jpn_Hira-swe", "jpn_Hira-tlh_Latn", "jpn_Hira-tur", "jpn_Hira-ukr", "jpn_Hira-vie", "jpn_Kana-rus", "jpn_Kana-spa", "kab-kab", "kab-rus", "kab-spa", "kat-rus", "kaz-rus", "kaz_Cyrl-rus", "khm-spa", "kor-rus", "kor-spa", "kor-zho", "kor_Hang-cmn_Hans", "kor_Hang-rus", "kor_Hang-spa", "kzj-msa", "kzj_Latn-zsm_Latn", "lad-lat", "lad-lfn", "lad-spa", "lad-yid", "lad_Latn-lfn_Latn", "lad_Latn-spa", "lad_Latn-yid", "lat-lat", "lat-lfn", "lat-nld", "lat-nor", "lat-pol", "lat-por", "lat-rus", "lat-spa", "lat-tlh", "lat-ukr", "lat-yid", "lat_Latn-lfn_Latn", "lat_Latn-por", "lav-rus", "lfn-por", "lfn-rus", "lfn-spa", "lfn-yid", "lfn_Cyrl-por", "lfn_Latn-por", "lfn_Latn-yid", "lit-pol", "lit-rus", "lit-spa", "lit-tur", "ltz-nld", "mkd-spa", "msa-msa", "msa-spa", "msa-zho", "nds-nld", "nds-por", "nds-rus", "nds-spa", "nld-cmn_Hant", "nld-nld", "nld-nor", "nld-pol", "nld-por", "nld-ron", "nld-rus", "nld-spa", "nld-toki", "nld-tur", "nld-ukr", "nld-zho", "nno-nob", "nob-nno", "nob-rus", "nob-spa", "nob-swe", "nor-nor", "nor-pol", "nor-por", "nor-rus", "nor-spa", "nor-swe", "nor-ukr", "nor-zho", "orv-ukr", "ota-tur", "pol-cmn_Hans", "pol-cmn_Hant", "pol-por", "pol-rus", "pol-spa", "pol-swe", "pol-tur", "pol-ukr", "pol-zho", "por-cmn_Hans", "por-cmn_Hant", "por-por", "por-ron", "por-rus", "por-spa", "por-swe", "por-tgl", "por-toki", "por-tur", "por-ukr", "por-zho", "ron-rus", "ron-spa", "ron-tur", "run-rus", "run-spa", "rus-cmn_Hans", "rus-cmn_Hant", "rus-rus", "rus-sah", "rus-slv", "rus-spa", "rus-swe", "rus-tat", "rus-tlh", "rus-toki", "rus-toki_Latn", "rus-tur", "rus-uig", "rus-uig_Arab", "rus-ukr", "rus-vie", "rus-xal", "rus-yue_Hans", "rus-zho", "slv-cmn_Hans", "slv-ukr", "slv-zho", "spa-cmn_Hans", "spa-cmn_Hant", "spa-spa", "spa-swe", "spa-tat", "spa-tgl", "spa-tlh", "spa-toki", "spa-tur", "spa-ukr", "spa-vie", "spa-yid", "spa-zho", "srp_Cyrl-rus", "srp_Cyrl-ukr", "srp_Latn-ita", "srp_Latn-nob", "srp_Latn-rus", "srp_Latn-ukr", "swe-cmn_Hans", "swe-cmn_Hant", "swe-swe", "swe-tur", "swe-zho", "tat-tur", "tat-vie", "tlh-yid", "tlh-zho", "tlh_Latn-cmn_Hans", "tlh_Latn-cmn_Hant", "tlh_Latn-yid", "tur-cmn_Hans", "tur-cmn_Hant", "tur-tur", "tur-uig", "tur-ukr", "tur-uzb", "tur-zho", "uig-zho", "uig_Arab-cmn_Hans", "uig_Arab-cmn_Hant", "ukr-cmn_Hans", "ukr-cmn_Hant", "ukr-ukr", "ukr-zho", "vie-cmn_Hans", "vie-vie", "vie-zho", "wuu-cmn_Hans", "yid-yid", "zho-zho", "zsm_Latn-ind", ]
+
+
+class tatoebaMTConfig(datasets.BuilderConfig):
+    """BuilderConfig for tatoeba-mt"""
+
+    def __init__(self, language_pair, release=_RELEASE, **kwargs):
+        super().__init__(**kwargs)
+        """
+
+        Args:
+            language_pair: language pair, you want to load
+            **kwargs: keyword arguments forwarded to super.
+        """
+        self.language_pair = language_pair
+        self.release = release
+
+
+class tatoebaMT(datasets.GeneratorBasedBuilder):
+    """tatoeba-mt is multilingual and covers a large number of language pairs."""
+
+    VERSION = datasets.Version("1.0.0")
+
+    BUILDER_CONFIG_CLASS = tatoebaMTConfig
+    BUILDER_CONFIGS = [
+        tatoebaMTConfig(name=pair, description=_DESCRIPTION, language_pair=pair, release=rel )
+        for pair in _LanguagePairs
+    ]
+
+    def _info(self):
+        return datasets.DatasetInfo(
+            description=_DESCRIPTION,
+            features=datasets.Features({
+                "sourceLang": datasets.Value("string"),
+                "targetlang": datasets.Value("string"),
+                "sourceString": datasets.Value("string"),
+                "targetString": datasets.Value("string")
+            }),
+            homepage="https://github.com/Helsinki-NLP/Tatoeba-Challenge/",
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+
+        lang_pair = self.config.language_pair
+        release = self.config.release
+        output = []
+
+        # testfile = os.path.join("test", release, f"tatoeba-test.{lang_pair}.tsv")
+        testfile = os.path.join("test", f"tatoeba-test.{lang_pair}.tsv")
+        if os.path.exists(testfile):
+            test = datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                # These kwargs will be passed to _generate_examples
+                gen_kwargs={
+                    "filepath": testfile,
+                },
+            )
+            output.append(test)
+
+        # devfile = os.path.join("dev", release, f"tatoeba-dev.{lang_pair}.tsv")
+        devfile = os.path.join("dev", f"tatoeba-dev.{lang_pair}.tsv")
+        if os.path.exists(devfile):
+            valid = datasets.SplitGenerator(
+                name=datasets.Split.VALIDATION,
+                # These kwargs will be passed to _generate_examples
+                gen_kwargs={
+                    "filepath": devfile,
+                },
+            )
+            output.append(valid)
+
+        return output
+
+    def _generate_examples(self, filepath):
+        """Yields examples."""
+        with open(filepath, encoding="utf-8") as f:
+            reader = csv.reader(f, delimiter="\t", quotechar='"')
+            for id_, row in enumerate(reader):
+                if id_ == 0:
+                    continue
+                yield id_, {
+                    "sourceLang": row[0],
+                    "targetlang": row[1],
+                    "sourceString": row[2],
+                    "targetString": row[3]
+                }
diff --git a/test/tatoeba-test.afr-deu.tsv b/test/tatoeba-test.afr-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5cf99c193219cbe85aff901db2aa2f73d20d05c9
--- /dev/null
+++ b/test/tatoeba-test.afr-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87f05c43cbd59ad12053d79b7d8d4cb3d2cd4f972167f2681f641f9b05e0bb75
+size 99941
diff --git a/test/tatoeba-test.afr-eng.tsv b/test/tatoeba-test.afr-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1b5834fa34f19e0f2bb2147f480ee3d53fd1d74e
--- /dev/null
+++ b/test/tatoeba-test.afr-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8305b1ec830d4a88109b757afa8bbbb890dc46f75d8faaa36d40f1b604d7d679
+size 96460
diff --git a/test/tatoeba-test.afr-epo.tsv b/test/tatoeba-test.afr-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e688bcbdc36dabcfccf8e52a19ac93f8c97d9ff0
--- /dev/null
+++ b/test/tatoeba-test.afr-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:603708f89e5a5d49d6a50fdc43fa2f9671e8895dacbef82c169312ed0fc346a1
+size 81498
diff --git a/test/tatoeba-test.afr-nld.tsv b/test/tatoeba-test.afr-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2fa70640e12f08004edd684431fb4fee5149909a
--- /dev/null
+++ b/test/tatoeba-test.afr-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03b09e5ec86aaa39d3a2b303c3bd10a953eb30cc853dfd697eeb0a6bbedcac46
+size 68060
diff --git a/test/tatoeba-test.afr-rus.tsv b/test/tatoeba-test.afr-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7b311287cdf22179e2a4bf8690d2d8bbc30bba63
--- /dev/null
+++ b/test/tatoeba-test.afr-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25463c43642132f4e524f540b18810a9d82a6b0ed35365f1461be81e7aacbc1a
+size 17527
diff --git a/test/tatoeba-test.afr-spa.tsv b/test/tatoeba-test.afr-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..964142b83f0b7fc962aac82ab0f4d4179844ef14
--- /dev/null
+++ b/test/tatoeba-test.afr-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7562769c10a519bf45ef8c4b2c2291f62cd87083eccdcff61ddffd1f2ae10f69
+size 29333
diff --git a/test/tatoeba-test.ain-fin.tsv b/test/tatoeba-test.ain-fin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9763994bd548d49c9686ccd4e8d4f786f544fe1f
--- /dev/null
+++ b/test/tatoeba-test.ain-fin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d41e018f361b94e7913206e8ba2a0b6d303950e335962be0c51145400bbe9f56
+size 14201
diff --git a/test/tatoeba-test.ara-ber.tsv b/test/tatoeba-test.ara-ber.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..88358cc3153fed320e6a6364a80d29aa385df476
--- /dev/null
+++ b/test/tatoeba-test.ara-ber.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27b7e9e2da20bbec414fc5ee0e38e55a64141bc7550c84296da591a89fe02b59
+size 209878
diff --git a/test/tatoeba-test.ara-ber_Latn.tsv b/test/tatoeba-test.ara-ber_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..67fd5d51f0dfaae2b071e005945e38e209e38afa
--- /dev/null
+++ b/test/tatoeba-test.ara-ber_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9361687ea6d78d93fb6a8877b7fc5fdc080fab4881156762f370d42e31b194b9
+size 208463
diff --git a/test/tatoeba-test.ara-deu.tsv b/test/tatoeba-test.ara-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a4106e822eb16ab9cacce6da96e0c5f6895fdf2a
--- /dev/null
+++ b/test/tatoeba-test.ara-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:723bae57b0ac605aa5c30f9badbb88bf7f47238c4c78fd3efa3e8b4c92a54c39
+size 102672
diff --git a/test/tatoeba-test.ara-ell.tsv b/test/tatoeba-test.ara-ell.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..759807a700cc6dd68d6577f6394a30e03cb1debb
--- /dev/null
+++ b/test/tatoeba-test.ara-ell.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d52a502f7747a2a3bb8c0112f39837cb452609db2f41b01b48c1a28d58084ef7
+size 32889
diff --git a/test/tatoeba-test.ara-eng.tsv b/test/tatoeba-test.ara-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6dfa8d7375e26ce4a13406f094e24e7679522cd4
--- /dev/null
+++ b/test/tatoeba-test.ara-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb5d17baf2c7cccc9a9b270f7620f316501d90b4235d655c073f2b9aaacc0e32
+size 938171
diff --git a/test/tatoeba-test.ara-epo.tsv b/test/tatoeba-test.ara-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6ec749e18e914e78008c6fa3437588b51bfd240f
--- /dev/null
+++ b/test/tatoeba-test.ara-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e583c0a92719e76f570b13258d2b5aad442525d48e58855ccc32e886c229698
+size 59071
diff --git a/test/tatoeba-test.ara-fra.tsv b/test/tatoeba-test.ara-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7f6fe92cb73a44dc62b639d5d889d28c9ad122a6
--- /dev/null
+++ b/test/tatoeba-test.ara-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba89f43e33800728e543dd1498f28b1d15e4c462d3d2e2c54e3568ed19163de1
+size 134171
diff --git a/test/tatoeba-test.ara-heb.tsv b/test/tatoeba-test.ara-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..040b209f466581c97d50d7ac79a7fbdb49da1b78
--- /dev/null
+++ b/test/tatoeba-test.ara-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75a25a4d35ae0ad030af3fbb03b76efb92249cea0c3dfea6465a16d855d41bcc
+size 110757
diff --git a/test/tatoeba-test.ara-ita.tsv b/test/tatoeba-test.ara-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e09b42942b25f0494f63ad76cffd00d7c40c3670
--- /dev/null
+++ b/test/tatoeba-test.ara-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bea76e35a95fd16002a77fa8ccee80e29691c471997161c03ab34d585515c7e9
+size 18780
diff --git a/test/tatoeba-test.ara-jpn.tsv b/test/tatoeba-test.ara-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..788636e57b5ec6e77bde693546710864bb4cac45
--- /dev/null
+++ b/test/tatoeba-test.ara-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd0f80de74586499c29a473c092d5cce4e0006f245b85fe6fd46878fea9abc48
+size 136439
diff --git a/test/tatoeba-test.ara-jpn_Hira.tsv b/test/tatoeba-test.ara-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c5fd0c7ded869e726f2db3d6f20981ca65ae699f
--- /dev/null
+++ b/test/tatoeba-test.ara-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0945006b3fbea6a5bcba4c1e71eaad03a587e24260bebd99d4d9178d99d14ef6
+size 111206
diff --git a/test/tatoeba-test.ara-pol.tsv b/test/tatoeba-test.ara-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ed63d007fd4d36ec1d4c197a4fabf1636cfda0ce
--- /dev/null
+++ b/test/tatoeba-test.ara-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c54d36c061aaca0f57801d954aae194c8655345c413820c90e3cc3006f652bb4
+size 16218
diff --git a/test/tatoeba-test.ara-rus.tsv b/test/tatoeba-test.ara-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..53af8ea14076fe092a53621568efc71e67a81c3d
--- /dev/null
+++ b/test/tatoeba-test.ara-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:261ce98449c34700d98c8bbb3b745cee3c6af3f06d5cdaef2c21d0a0f1484252
+size 366839
diff --git a/test/tatoeba-test.ara-spa.tsv b/test/tatoeba-test.ara-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..17d24bfadac7127184d99c5f7ef18be17f4c8eb3
--- /dev/null
+++ b/test/tatoeba-test.ara-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0d63254af3241f89b2aea353abe222a17388165209cd88e4a7f0e6eaccd3ce6
+size 118851
diff --git a/test/tatoeba-test.ara-tur.tsv b/test/tatoeba-test.ara-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..77e998a506790acbd59202ee705dc285564abe8c
--- /dev/null
+++ b/test/tatoeba-test.ara-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0834a43f79c45db532f137cbdef64193eb64660f36dca684fac2ff81007605a9
+size 109267
diff --git a/test/tatoeba-test.arq-eng.tsv b/test/tatoeba-test.arq-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fc4927dfdea12b7429a79f4094f919352e2043ad
--- /dev/null
+++ b/test/tatoeba-test.arq-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:638dd22cc3c2f08b58c47c34d73bb216a20aa539c8b61877af510ad425c45a0e
+size 36058
diff --git a/test/tatoeba-test.avk-fra.tsv b/test/tatoeba-test.avk-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4b305b879b95c7dc897a01b4b398dbd5b85c4ef3
--- /dev/null
+++ b/test/tatoeba-test.avk-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a73b6c86e7555ff158b1f28ea022907fbe35bd08a80b002f47ca19de846b6d8
+size 107793
diff --git a/test/tatoeba-test.avk-spa.tsv b/test/tatoeba-test.avk-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..84d867c6213c95e13a6e11cd984d897e7839aa4b
--- /dev/null
+++ b/test/tatoeba-test.avk-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:518b8082a0ee3bd8a1b06fd1814d2b74dbf0dd72045d40297dfb21a47bd7c522
+size 18491
diff --git a/test/tatoeba-test.awa-eng.tsv b/test/tatoeba-test.awa-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5a1e1fa79d854a2767c740ecb48914eb6ccb948b
--- /dev/null
+++ b/test/tatoeba-test.awa-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e737c6c2f61129f48f34cfb9b567f53f4c5351ee150be379cd676a0bc8fce06f
+size 20696
diff --git a/test/tatoeba-test.aze-eng.tsv b/test/tatoeba-test.aze-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d6c5d5209fe250b9810973f529b64dd068f1ea1f
--- /dev/null
+++ b/test/tatoeba-test.aze-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ae2af557c7a76c0eac2a7f9e830142aba73c4ac7f6c402c8ec9d8ab9fd8b87a
+size 185471
diff --git a/test/tatoeba-test.aze-spa.tsv b/test/tatoeba-test.aze-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..51395118a3682d3bb7e5f3013901e73b035b7eb4
--- /dev/null
+++ b/test/tatoeba-test.aze-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:166fb1a2d6ce324c222e5c90c0e945a72486d68b8ba3fd5642006891b67ff703
+size 13994
diff --git a/test/tatoeba-test.aze-tur.tsv b/test/tatoeba-test.aze-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d4ccaa7dc44eccd278263d6017a746bad8cb8cb5
--- /dev/null
+++ b/test/tatoeba-test.aze-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ad8489e9c8d5a5a4c301e976593dd9eb696e34e9c08bb97f7f475e153560778
+size 81567
diff --git a/test/tatoeba-test.aze_Latn-tur.tsv b/test/tatoeba-test.aze_Latn-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..35fead7c0d743514b393e9179d3112a91bff4b71
--- /dev/null
+++ b/test/tatoeba-test.aze_Latn-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1bcd8bcb4e5834c541a1fdd7542a3bccbb6501bb1d945a0a672231dd3dad6b7
+size 80884
diff --git a/test/tatoeba-test.bel-deu.tsv b/test/tatoeba-test.bel-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5dd52de861133597057e48b1aef0d0d7ae58322d
--- /dev/null
+++ b/test/tatoeba-test.bel-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05da5604882834d30594e1701b586bc25abd05a6876e4bd82d08facb2a95e658
+size 56902
diff --git a/test/tatoeba-test.bel-eng.tsv b/test/tatoeba-test.bel-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c2547e30e86b92e9e99b33fcbdb7d740da9e62e8
--- /dev/null
+++ b/test/tatoeba-test.bel-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:347518dc4a18c95120116e43f3900afbcdd550e0b89f749da0864c3b29708bd0
+size 244579
diff --git a/test/tatoeba-test.bel-epo.tsv b/test/tatoeba-test.bel-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3a50d9028ba015d05bc22d4c8349fce2066c6aca
--- /dev/null
+++ b/test/tatoeba-test.bel-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09d5ddc475c1bfaecba706063a90d568adcb30ec777c2d8238a7c8ac09be6160
+size 75827
diff --git a/test/tatoeba-test.bel-fra.tsv b/test/tatoeba-test.bel-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a8e5020863b115cb54617bdc5e5d34335a545fa3
--- /dev/null
+++ b/test/tatoeba-test.bel-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f916f085f92eed6940ce97915ec41b067e204dd331937ebeaf981d8ba9e6824
+size 26819
diff --git a/test/tatoeba-test.bel-ita.tsv b/test/tatoeba-test.bel-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cb525586015c6c5994071f7b9eaad4d9841300c9
--- /dev/null
+++ b/test/tatoeba-test.bel-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cc65aa7095f62132ab16a1b91829dfe2183a561c54204895f9870995badb00e
+size 22431
diff --git a/test/tatoeba-test.bel-lat.tsv b/test/tatoeba-test.bel-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8fa245eb33dad0f4b0d3e149bd7c5a3484c77c81
--- /dev/null
+++ b/test/tatoeba-test.bel-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64342700669558ac3fd5808ed893a661bee13522a70814aafe7460f8a11fdaa2
+size 19339
diff --git a/test/tatoeba-test.bel-nld.tsv b/test/tatoeba-test.bel-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..439bdebd0a99ec2919b99c47705a6fb03027c805
--- /dev/null
+++ b/test/tatoeba-test.bel-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:480be792390d368ae004044d1eda25fb6470022d57d9c7b803b7a2cc92d83289
+size 62683
diff --git a/test/tatoeba-test.bel-pol.tsv b/test/tatoeba-test.bel-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0a00d9be08b8374d4fffd820918d2ab0218f1dea
--- /dev/null
+++ b/test/tatoeba-test.bel-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37311a6ef6b6da461fa594f0ed074af938c9b15d33888199369aeb67f6e17cf2
+size 26236
diff --git a/test/tatoeba-test.bel-rus.tsv b/test/tatoeba-test.bel-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d14637b54186be6700199c35dc107eb5f1286c06
--- /dev/null
+++ b/test/tatoeba-test.bel-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf5b319c3fabc9b9d1cdefc8a2721579b0596b7e1bbc2b0cabdefe9465bc3b3e
+size 365022
diff --git a/test/tatoeba-test.bel-spa.tsv b/test/tatoeba-test.bel-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bf192c81359d58824d884587b0054cb554716fdb
--- /dev/null
+++ b/test/tatoeba-test.bel-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f89dfcd10d46fb94041cfa184985122e7888ce272f074b420968fd4f92e4eb1e
+size 18897
diff --git a/test/tatoeba-test.bel-ukr.tsv b/test/tatoeba-test.bel-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1fcbdea5f7ce86d7e02f7457feeffc6df7228eb0
--- /dev/null
+++ b/test/tatoeba-test.bel-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2912ec386f7806a1cab96f70a14f6c29e2d0f1c072231987cb0054bb597f1528
+size 284639
diff --git a/test/tatoeba-test.bel-zho.tsv b/test/tatoeba-test.bel-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..860f02a5464ee4377cb5debf3cc0f7b724d8c32e
--- /dev/null
+++ b/test/tatoeba-test.bel-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:799d352c5f7717e29dd6111e6f03dde670836313a55a5425db427ce4c096bc2d
+size 29413
diff --git a/test/tatoeba-test.ben-eng.tsv b/test/tatoeba-test.ben-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1f33ae1830564fbf92014c82dce07626a17b8d91
--- /dev/null
+++ b/test/tatoeba-test.ben-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41b4d381439cfb210993bc782deeb7e34850414baae63e00b4ce9e83ab08204f
+size 234586
diff --git a/test/tatoeba-test.ber-deu.tsv b/test/tatoeba-test.ber-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..222d5e53b8ac5e69a6ffec5e0c5c6a41fb9bfb48
--- /dev/null
+++ b/test/tatoeba-test.ber-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e89cc2fe0b10acb01a8400b4c271657060cbfd058eb7f55c68c7a90ac12a4885
+size 56019
diff --git a/test/tatoeba-test.ber-eng.tsv b/test/tatoeba-test.ber-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a62fe0f6196df30d89aa9c037af181b3b774567c
--- /dev/null
+++ b/test/tatoeba-test.ber-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0f86a398c4a4b840fcfe39c3fc5bd86764fec9b6925ef413ba92470b5001348
+size 896825
diff --git a/test/tatoeba-test.ber-epo.tsv b/test/tatoeba-test.ber-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fcc0e9d006bb40cefc252b11b09c45108625ff02
--- /dev/null
+++ b/test/tatoeba-test.ber-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55bab4b0bd2c5ed8a526606680d7027580e92557758da6aeed326ea74fcb5e10
+size 133156
diff --git a/test/tatoeba-test.ber-fra.tsv b/test/tatoeba-test.ber-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4c0a21b9af5c3a88458b63580feba20925551e5d
--- /dev/null
+++ b/test/tatoeba-test.ber-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26ee91480798f01647fdd14ec0d813b0ff4af7841872d24cc1327952f2043edd
+size 758124
diff --git a/test/tatoeba-test.ber-spa.tsv b/test/tatoeba-test.ber-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..df6e398ae44e6bef6b4fa9778a778a0b0fe216fc
--- /dev/null
+++ b/test/tatoeba-test.ber-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f9d16c75999005aeb7c3d06f78741269fb742ac6f60577aedce6fc1cf924198
+size 807626
diff --git a/test/tatoeba-test.ber_Latn-deu.tsv b/test/tatoeba-test.ber_Latn-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..72361cbd226ae055a2f921047ccca63436874138
--- /dev/null
+++ b/test/tatoeba-test.ber_Latn-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0258735167aa790032a6641eb474b7761977115b1925b7933366ac226597bbe
+size 55766
diff --git a/test/tatoeba-test.ber_Latn-eng.tsv b/test/tatoeba-test.ber_Latn-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..79713660d6846a1d80c4aeb3a2e5174db1700987
--- /dev/null
+++ b/test/tatoeba-test.ber_Latn-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:308879d76731a1cdb5fdd3c19bf0bccb14a178bc960bb625ad0fc29d6dfa381b
+size 896660
diff --git a/test/tatoeba-test.ber_Latn-epo.tsv b/test/tatoeba-test.ber_Latn-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b43fbbc93ebc58188a2ceb637a32014fdf0b0fae
--- /dev/null
+++ b/test/tatoeba-test.ber_Latn-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7a0fcff23cb915c8435dd4f0efeb70ee058e714ff774f5db5790dd3983867d1
+size 133008
diff --git a/test/tatoeba-test.ber_Latn-fra.tsv b/test/tatoeba-test.ber_Latn-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ebe64ee1a41713831b528628d8eb35f1f89ce9d3
--- /dev/null
+++ b/test/tatoeba-test.ber_Latn-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:468f297e232ded64e180f95d9e40a007a91d6a0478a57993487bdf888b34ac37
+size 757783
diff --git a/test/tatoeba-test.bre-eng.tsv b/test/tatoeba-test.bre-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3dbdc9ce8cb72e3980c7a937da182da5cfc38183
--- /dev/null
+++ b/test/tatoeba-test.bre-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ce0ac3199b819f77ebc637c77363705948c1ccd4a69f4fabbcd5f31cf78f89f
+size 21410
diff --git a/test/tatoeba-test.bre-fra.tsv b/test/tatoeba-test.bre-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3057d9ef8e3e6b8e1c9a9fb2af25ad4348375133
--- /dev/null
+++ b/test/tatoeba-test.bre-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c3a6036891e095164a11c3b572e922c816c50ea4fe40bc9b2a0f930edac3393
+size 145228
diff --git a/test/tatoeba-test.bua-rus.tsv b/test/tatoeba-test.bua-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0c88aa42b19548d6ddc535736a61e3bf401f4f5c
--- /dev/null
+++ b/test/tatoeba-test.bua-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84bcea757be551230fa91c0a1cf52b60b60177716b25ba25063f916897867121
+size 74438
diff --git a/test/tatoeba-test.bua_Cyrl-rus.tsv b/test/tatoeba-test.bua_Cyrl-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9afc761c9266003042d6fa1d6aeb23292e05659c
--- /dev/null
+++ b/test/tatoeba-test.bua_Cyrl-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5f3dfc3556c5f759dbfcf0eb2afc4fc1d3dddb9d323aaf34f93894e50c50c98
+size 74416
diff --git a/test/tatoeba-test.bul-bul.tsv b/test/tatoeba-test.bul-bul.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0e88790893213471168e2a915323defc88606751
--- /dev/null
+++ b/test/tatoeba-test.bul-bul.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce6fbe4f8290f2e41c290faae731f34220408c75d8b53c80022aced6bd54041d
+size 23925
diff --git a/test/tatoeba-test.bul-cmn_Hans.tsv b/test/tatoeba-test.bul-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9609a7a10b705deb663e83d82c3d39193f41d474
--- /dev/null
+++ b/test/tatoeba-test.bul-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63dcfc5a6628ee13554dcd5bb67ff1e9cc1a2f608dfbbefd56e3540d9b19c06d
+size 31599
diff --git a/test/tatoeba-test.bul-deu.tsv b/test/tatoeba-test.bul-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..efd039831181269cb79e1ef8046763e3a286b5b8
--- /dev/null
+++ b/test/tatoeba-test.bul-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a654bb4bb87e07b9963d8cd20fcbe85d5f9e4f108087700704d912747fde0205
+size 30294
diff --git a/test/tatoeba-test.bul-eng.tsv b/test/tatoeba-test.bul-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a48876122d14e6aa9a813f21a4639bedaa100641
--- /dev/null
+++ b/test/tatoeba-test.bul-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea5ad1a26ab5dce6ebde36efb273fb1c9d84966ece493fb1b3c0ad37505695b0
+size 964374
diff --git a/test/tatoeba-test.bul-epo.tsv b/test/tatoeba-test.bul-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e2a95c0dcb9415b2cba235a93f41c7f4e2697844
--- /dev/null
+++ b/test/tatoeba-test.bul-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:682b94a452379f6c113e7ba885bee08d9d98f1e67ea11d355dd085bfaba22185
+size 56818
diff --git a/test/tatoeba-test.bul-fra.tsv b/test/tatoeba-test.bul-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a6f3fb67b7f68b97834acb7e110dfdc8ed1aed3a
--- /dev/null
+++ b/test/tatoeba-test.bul-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6920d87374e2a65ccb6fa10d4ef5a280c35ec476f695254e7f8517953cb577cc
+size 48703
diff --git a/test/tatoeba-test.bul-ita.tsv b/test/tatoeba-test.bul-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..804ef2a5a66c60a78d889c94220229f70f321283
--- /dev/null
+++ b/test/tatoeba-test.bul-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3cf9d02513a6f98444afbc92122bf96f89f3c9452bbb9a12aa0119e4ef6d433c
+size 241262
diff --git a/test/tatoeba-test.bul-jpn.tsv b/test/tatoeba-test.bul-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..015591c52af273d01939a04e2fd9e25a0745d2ca
--- /dev/null
+++ b/test/tatoeba-test.bul-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d306329088d297c8d2c0ec4474a46c29a78094f64cc2f9ff415f1f9bd4d615ec
+size 39835
diff --git a/test/tatoeba-test.bul-jpn_Hira.tsv b/test/tatoeba-test.bul-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2ddb5c19203a49906df36cc6de6a3a3c5fda8579
--- /dev/null
+++ b/test/tatoeba-test.bul-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f147758d82c193412f803d56e88433feefad870effd628b5968b038375e9f70
+size 33801
diff --git a/test/tatoeba-test.bul-rus.tsv b/test/tatoeba-test.bul-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..39e17f0c9998e437bfee8a365234648b262e795b
--- /dev/null
+++ b/test/tatoeba-test.bul-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8bbcd6a8532d48bea610c4705f62a8d47f94c180e6dacb04d455b3e90cc837d1
+size 142585
diff --git a/test/tatoeba-test.bul-spa.tsv b/test/tatoeba-test.bul-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..77c84088eb60fc8707a86279db1201f21d06a84c
--- /dev/null
+++ b/test/tatoeba-test.bul-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ce69109109a1bada453896978290de3ed4aadb2425d81a6165ee72dc0afc134
+size 24507
diff --git a/test/tatoeba-test.bul-tur.tsv b/test/tatoeba-test.bul-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e61bc278bc426fa132c00395c9a67fc4f54bd223
--- /dev/null
+++ b/test/tatoeba-test.bul-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84dfe392c36fa6f0f33e434c76841f30ae6bcbf1ac4b266757633a84d0a76cad
+size 88651
diff --git a/test/tatoeba-test.bul-ukr.tsv b/test/tatoeba-test.bul-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..38541a133c9ea330aae13f7be2b16784e854d87c
--- /dev/null
+++ b/test/tatoeba-test.bul-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c86e30122a8d7eff91b457ac13a9e732169cd240e2d005744ab63b930630d37e
+size 88243
diff --git a/test/tatoeba-test.bul-zho.tsv b/test/tatoeba-test.bul-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3edcf30cfed41f8cda41de635e20229f9e94800a
--- /dev/null
+++ b/test/tatoeba-test.bul-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1d1da6ecdd756df198518e4ec13734f18a474e9e77bad52728e6e9475a0fdc1
+size 45135
diff --git a/test/tatoeba-test.cat-deu.tsv b/test/tatoeba-test.cat-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..39a53a8577061af487b8fe18639322975f388e76
--- /dev/null
+++ b/test/tatoeba-test.cat-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdd2538ad6fc2ad996240e67508e52dc861e364a32726e1830b7a58e1d9d58bc
+size 59959
diff --git a/test/tatoeba-test.cat-eng.tsv b/test/tatoeba-test.cat-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6c4fa02174e236b13d4711bdf2979123e60ace52
--- /dev/null
+++ b/test/tatoeba-test.cat-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1947312f2f9702aa9a56d4544e9d81946029708341d21952cfddac68f94ed19c
+size 126706
diff --git a/test/tatoeba-test.cat-epo.tsv b/test/tatoeba-test.cat-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a82aba7f8bf95b3c9a317ad5ca6b57a1d9ac2dd9
--- /dev/null
+++ b/test/tatoeba-test.cat-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2319763dd64e90e7a8bc062149a43bdc2c5b6ee33f7e1fba5a1680ce8a692d0
+size 61422
diff --git a/test/tatoeba-test.cat-fra.tsv b/test/tatoeba-test.cat-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c620daa630966ecaf4f8e5cbe9fb6bad6444ecde
--- /dev/null
+++ b/test/tatoeba-test.cat-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:188aaed4b953e581ed2e4777d83075c1404d0bacc919f8da95c273bf756f0da8
+size 58832
diff --git a/test/tatoeba-test.cat-ita.tsv b/test/tatoeba-test.cat-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c42c140adcce0dc86b1580d028ac1d1ac254d567
--- /dev/null
+++ b/test/tatoeba-test.cat-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2be9637408ade652d4ee0df07d1fbe8d9c83284b85802583f0e908fc45a1389f
+size 22469
diff --git a/test/tatoeba-test.cat-nld.tsv b/test/tatoeba-test.cat-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..84b0a96d3db6d1c8e7d821d924feb286aa8756bd
--- /dev/null
+++ b/test/tatoeba-test.cat-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05776b49cadc5e26e15328bd70a4c914581c7a8659f89c1028e7500233177b49
+size 41913
diff --git a/test/tatoeba-test.cat-por.tsv b/test/tatoeba-test.cat-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..802dc8bacbb5b6660f54d16e9d8e87ce4735eca3
--- /dev/null
+++ b/test/tatoeba-test.cat-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:233896fc35a7648e5bae45b8eb2993b9380f4b7037962b7c828c135a249e9729
+size 65694
diff --git a/test/tatoeba-test.cat-spa.tsv b/test/tatoeba-test.cat-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4e14a336123e72af730370d0947e3c310dc86959
--- /dev/null
+++ b/test/tatoeba-test.cat-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b7640e34ce2a2f219abaa2897daeb03f3ae87024bd044319f3e6ff67032611f
+size 131262
diff --git a/test/tatoeba-test.cat-ukr.tsv b/test/tatoeba-test.cat-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..51a952669998d4cbc7628b55ac8d923bed1f36d8
--- /dev/null
+++ b/test/tatoeba-test.cat-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:391c899c16b73a152b35ce13890a683a1eb3a7d2669b7c75fd13f1a26bf9fea8
+size 35263
diff --git a/test/tatoeba-test.cbk-eng.tsv b/test/tatoeba-test.cbk-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5284eb574721cb75c268035b099dd765af70d0c4
--- /dev/null
+++ b/test/tatoeba-test.cbk-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9f9694a302c45e8ff4cc3f9d37180a7c1a30434b3d4e11403835790d210254c
+size 110361
diff --git a/test/tatoeba-test.ceb-deu.tsv b/test/tatoeba-test.ceb-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..975a70aca242883ba8cd1bf614cc3043d61a7d9f
--- /dev/null
+++ b/test/tatoeba-test.ceb-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ffd69377f4dc0ec62a2e93d3c6887573ebf2caaa9720c7e2834c3fded099cccf
+size 76229
diff --git a/test/tatoeba-test.ceb-eng.tsv b/test/tatoeba-test.ceb-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..612b38a1bafcc5d83d87b2f6511479665185533e
--- /dev/null
+++ b/test/tatoeba-test.ceb-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ed25df67b5b50aed2d68b51a5cd1446e41fc80eebb94d910a9c4d02816db3ca
+size 23173
diff --git a/test/tatoeba-test.ces-deu.tsv b/test/tatoeba-test.ces-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..544553f551cff20ce3ef7acb68de4d922e40d662
--- /dev/null
+++ b/test/tatoeba-test.ces-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4300c176cf9bcafd9a05c8374a3b61711880420ec86d85f1f4b377132f7f8ae
+size 282537
diff --git a/test/tatoeba-test.ces-eng.tsv b/test/tatoeba-test.ces-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..53a20b2a3db98379fd145344f69415bdc3bebe0a
--- /dev/null
+++ b/test/tatoeba-test.ces-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33a2afe417be86dc16a85e8308c0c55912f422f50c57bf4f297b73284db5ce6a
+size 1043172
diff --git a/test/tatoeba-test.ces-epo.tsv b/test/tatoeba-test.ces-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..17356f1b7ef8843bca9f8007e2c0b2985e57a2b9
--- /dev/null
+++ b/test/tatoeba-test.ces-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3cae4c4eff60ca350bf98a8ffb36efb38fa542dbe397a6586081ff8d5da32bce
+size 262803
diff --git a/test/tatoeba-test.ces-fra.tsv b/test/tatoeba-test.ces-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dc094ecc4b28b9eb5ad04111e02e4165d9bcd8ad
--- /dev/null
+++ b/test/tatoeba-test.ces-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7eab2d2d424a3f9d34559e132f6351a2b4513363c2316e66c10ee32f1fecea15
+size 34150
diff --git a/test/tatoeba-test.ces-hun.tsv b/test/tatoeba-test.ces-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2cb8351150df88a92ada460d9f484fee66b29e41
--- /dev/null
+++ b/test/tatoeba-test.ces-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:525facc65c625f258d2b9585c7c6c86539863e514bd158f0f58a9c606b1c8931
+size 121638
diff --git a/test/tatoeba-test.ces-ita.tsv b/test/tatoeba-test.ces-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9acb30c58a7b8b5f39b1b2580bd72d956270a893
--- /dev/null
+++ b/test/tatoeba-test.ces-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94231c6b4afe56034eb21900fcdefe04b637c48c5fa1b43b3e88a8f060432176
+size 85180
diff --git a/test/tatoeba-test.ces-lat.tsv b/test/tatoeba-test.ces-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8230ecfab38a37d84107991bbc9a794cd588d588
--- /dev/null
+++ b/test/tatoeba-test.ces-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1952623a2c96a774164d66bfebd53f8ce67eb33264aa15dcb176c758a20faeb1
+size 22590
diff --git a/test/tatoeba-test.ces-pol.tsv b/test/tatoeba-test.ces-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9c5a3ccafdee074bbc58d3edf6119d3b23940d21
--- /dev/null
+++ b/test/tatoeba-test.ces-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03e934797a0ba391a17f74a421ff9cd8a2752fe5639643ee093c83c7c1ea13e4
+size 40371
diff --git a/test/tatoeba-test.ces-rus.tsv b/test/tatoeba-test.ces-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b7343140eb1abcf635940e192492aa1f75485e77
--- /dev/null
+++ b/test/tatoeba-test.ces-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:476be95c39053bbc2d97185b1fd9e8bb0881386c0ccd1acaf5ed6fa90aa2d043
+size 254127
diff --git a/test/tatoeba-test.ces-slv.tsv b/test/tatoeba-test.ces-slv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b60e209bce84ad972782933dc36ae976a0b3a4df
--- /dev/null
+++ b/test/tatoeba-test.ces-slv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68f9cd69c0c4c6cd799754564f204d5c2c417a2d3ccf5ba3d413b3df80d3b8ca
+size 28804
diff --git a/test/tatoeba-test.ces-spa.tsv b/test/tatoeba-test.ces-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0ef70762271174343f9be725b65f86b9316b2aa5
--- /dev/null
+++ b/test/tatoeba-test.ces-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fc62493a0600314430712da13b8c9f7277e3e887ab47f70875a3a4b3e35559b
+size 136343
diff --git a/test/tatoeba-test.ces-ukr.tsv b/test/tatoeba-test.ces-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c130ef6359b95a6dcdde47fbcfebb1dbfa55b94f
--- /dev/null
+++ b/test/tatoeba-test.ces-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:43019313403e7f6e7ec56f90379ea97f8a51d2f641d5563b58daa5750810401d
+size 127271
diff --git a/test/tatoeba-test.cha-eng.tsv b/test/tatoeba-test.cha-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1f069a06b719c2c3ee41508db5123cfe82dcf05f
--- /dev/null
+++ b/test/tatoeba-test.cha-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec8e1112f0be5e54db0ab10440827b6e6ce17cb4233f0400819045230a8bbd2d
+size 11634
diff --git a/test/tatoeba-test.chm-rus.tsv b/test/tatoeba-test.chm-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..065ddf77faabfd6f8157fc5b62b3e25db956106c
--- /dev/null
+++ b/test/tatoeba-test.chm-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:996e5f44bbf74c26f8ebc27c816d5582078889ede4a8703c48a4e26d84031f8b
+size 381966
diff --git a/test/tatoeba-test.chv-eng.tsv b/test/tatoeba-test.chv-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9d672bca3c167d022b6a0956d1818064a94b4843
--- /dev/null
+++ b/test/tatoeba-test.chv-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1ef59000704ddc67dbd6a47e63a840972cae2f6bfca94089f8dced408a8ee01
+size 28676
diff --git a/test/tatoeba-test.chv-rus.tsv b/test/tatoeba-test.chv-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2d293acda124cb26c7babfb3112443d63445b9ed
--- /dev/null
+++ b/test/tatoeba-test.chv-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f72a43b26fa60a9c404cbf0b839104de9c5cd58249fbe059fd83798abb46d79
+size 40925
diff --git a/test/tatoeba-test.chv-tur.tsv b/test/tatoeba-test.chv-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..39b5ca3d6e98ded199f5fc2c703ed5f8af145708
--- /dev/null
+++ b/test/tatoeba-test.chv-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bbc9cb90d468f962fb0d397f57c947730293b06063bec823a2ee3801594102a
+size 44708
diff --git a/test/tatoeba-test.cmn_Hans-wuu.tsv b/test/tatoeba-test.cmn_Hans-wuu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dd7977183bfd81a99b542e2cbde798c70d6e6296
--- /dev/null
+++ b/test/tatoeba-test.cmn_Hans-wuu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:111377d98e13f0d1b7ecb4794720385812d09d7525a78f11338fbae72b7c66cf
+size 73864
diff --git a/test/tatoeba-test.cor-deu.tsv b/test/tatoeba-test.cor-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a33d9fdd3715e3a29c9ec2f82e3f75e2a4a52aab
--- /dev/null
+++ b/test/tatoeba-test.cor-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78fb6702707dbee99f2ad667cc768b1a460cb6fab15ba53c8103e49cecddd1f8
+size 47723
diff --git a/test/tatoeba-test.cor-eng.tsv b/test/tatoeba-test.cor-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..62d9b37f29e1cf6b17f37d6821b07e86ce61e730
--- /dev/null
+++ b/test/tatoeba-test.cor-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ef54148577ddf97072860ce3593fbd4ebaa195979ee911e617b44041c9564c3
+size 163747
diff --git a/test/tatoeba-test.cor-epo.tsv b/test/tatoeba-test.cor-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..836e758707484f754c58e60869fd7a0ae1cc1e45
--- /dev/null
+++ b/test/tatoeba-test.cor-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0b32bc7a3d1de09d909f96f95de4a721504fe7e36db8211a774263bd9609679
+size 36954
diff --git a/test/tatoeba-test.cor-fra.tsv b/test/tatoeba-test.cor-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..927a0b9b0403c8a28c9fcec09aa8e3cf8f958461
--- /dev/null
+++ b/test/tatoeba-test.cor-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdc5a453cf255267adda7ae6226fa67716c73a63136c10baa347be780db120ce
+size 31711
diff --git a/test/tatoeba-test.cor-ita.tsv b/test/tatoeba-test.cor-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..67df2343be21ca5c96bc746d58a9e8d173b5bd59
--- /dev/null
+++ b/test/tatoeba-test.cor-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b35e336ccdd32635126376a22cf1f0065d239716e5becd990b247f960dbb18d0
+size 14528
diff --git a/test/tatoeba-test.cor-rus.tsv b/test/tatoeba-test.cor-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..338ed4247a2ab74b7b4e2a03d3fd97d1f20ed6c3
--- /dev/null
+++ b/test/tatoeba-test.cor-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cccdff703fda0a3fa8eec8cc0e53744cd99e488b7461b280790a4530e2480f36
+size 14472
diff --git a/test/tatoeba-test.cor-spa.tsv b/test/tatoeba-test.cor-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0739c035b0df2f9a2f59a96f8ad2e186d5e72f91
--- /dev/null
+++ b/test/tatoeba-test.cor-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38f03ec48c9f65edb0c5589ec6305dfe5e6843ae309e4abf73381c94340448fc
+size 11511
diff --git a/test/tatoeba-test.crh-tur.tsv b/test/tatoeba-test.crh-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5e4b85d5fd84962936bd46db9b845cb4927a4d03
--- /dev/null
+++ b/test/tatoeba-test.crh-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc2497ff5a5484643c928b4d45ceade418f57c9723c3f11e39e314c003c2721c
+size 14520
diff --git a/test/tatoeba-test.cym-eng.tsv b/test/tatoeba-test.cym-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..74be988fd985242184d7ba5f5922aabee8841a9e
--- /dev/null
+++ b/test/tatoeba-test.cym-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc5c6b300a7a5d9164beb1d226aaa8ceef66530704cc81fdb872cd73acb7f6de
+size 57737
diff --git a/test/tatoeba-test.dan-dan.tsv b/test/tatoeba-test.dan-dan.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5f666926404cd0044b96a89a45fe7e449b21a0f6
--- /dev/null
+++ b/test/tatoeba-test.dan-dan.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc1cea5b84822bd152763330171fe2ea0c305c5e9b57629cb9684d44c2b7b886
+size 18075
diff --git a/test/tatoeba-test.dan-deu.tsv b/test/tatoeba-test.dan-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..caaf06171cdd7c2e6c2ed4496cdea51b753e3bc2
--- /dev/null
+++ b/test/tatoeba-test.dan-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68d3ae2c7a67475c2d226631d6e0ed833992130708874bdb6c210d698f136fd8
+size 819673
diff --git a/test/tatoeba-test.dan-eng.tsv b/test/tatoeba-test.dan-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..251b52d3b57d44fdf1f0cea43a6dc623fe4ce042
--- /dev/null
+++ b/test/tatoeba-test.dan-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2039e4abf65151a0adb2343713c4135c5ed6475f5c12ca991d0079a0eea77e92
+size 807981
diff --git a/test/tatoeba-test.dan-epo.tsv b/test/tatoeba-test.dan-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0e334c64d0a9d5e5ac9008424409f5124b7ee62a
--- /dev/null
+++ b/test/tatoeba-test.dan-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8909a7167d4d90e64fe5a2678a87d803e8f06089da8fe49a2bfed65a45fb3484
+size 843351
diff --git a/test/tatoeba-test.dan-fin.tsv b/test/tatoeba-test.dan-fin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c19a58bffc7ba877325e6068e0989a0ce90b7b5c
--- /dev/null
+++ b/test/tatoeba-test.dan-fin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:456f559e88c7a8e79995349b8110eda89232351166119765f3d00f35b2b2bf21
+size 178869
diff --git a/test/tatoeba-test.dan-fra.tsv b/test/tatoeba-test.dan-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9da6b5f380072aeb14e5052c27f9ae659dc068b6
--- /dev/null
+++ b/test/tatoeba-test.dan-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b2e2b949d52458a2f92ffc7aed61fb51f38bfa28b74fc922247afa10eba6273
+size 120949
diff --git a/test/tatoeba-test.dan-ita.tsv b/test/tatoeba-test.dan-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ed42f340d6ffc333a164054a68b67c0e0a073ad2
--- /dev/null
+++ b/test/tatoeba-test.dan-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bc348f98430dc2630cac8ecd30a3dfd8333955e9df142dc6bcac5c11732c28e
+size 23819
diff --git a/test/tatoeba-test.dan-jpn.tsv b/test/tatoeba-test.dan-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1f9bc7885468ae110a2e25f6eb39e511f3b3e2a8
--- /dev/null
+++ b/test/tatoeba-test.dan-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d7cff952c7e8b460628a8f714a1d40e7749313051032b4ef4331c0042bce587
+size 74725
diff --git a/test/tatoeba-test.dan-jpn_Hira.tsv b/test/tatoeba-test.dan-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..19c47a95b5c4a3b112cfa1e6d19f7e818bc31207
--- /dev/null
+++ b/test/tatoeba-test.dan-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:080f9e20a668b2ba866c08ad9c56a253ab6a8767cae667af92d8316a1fb2ce01
+size 59987
diff --git a/test/tatoeba-test.dan-nld.tsv b/test/tatoeba-test.dan-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3618f331edc3b7e2087d6a249d8f191c187f513a
--- /dev/null
+++ b/test/tatoeba-test.dan-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64df7dbec50633cb010ed2b875f3635e949fcff5ad4ea28b05f681813fdbc35c
+size 109243
diff --git a/test/tatoeba-test.dan-nob.tsv b/test/tatoeba-test.dan-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0f2f5c43d9c73d39159e27c07b80157194a3bdcd
--- /dev/null
+++ b/test/tatoeba-test.dan-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ddcf474b1e544ca165e634629183017b00351294ff7f2f822af861977fce12f
+size 101643
diff --git a/test/tatoeba-test.dan-nor.tsv b/test/tatoeba-test.dan-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7cc51e178fed9943e7e983996c6bbf8c7ffa551c
--- /dev/null
+++ b/test/tatoeba-test.dan-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d4632a68c59bad9364a62fafcccd4bbdbea0c284ccd33f007a93f21baecdbaa
+size 102342
diff --git a/test/tatoeba-test.dan-por.tsv b/test/tatoeba-test.dan-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..65f21c2dc364187f04fdada4291f4ff1b5f899d9
--- /dev/null
+++ b/test/tatoeba-test.dan-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0139632fa3088ae245b768e36b8639e5668f7f49e594a9842dcfccfb7dd5389
+size 54721
diff --git a/test/tatoeba-test.dan-rus.tsv b/test/tatoeba-test.dan-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a60daeb5a4c1e161b26520ad5a4b97f37cfef991
--- /dev/null
+++ b/test/tatoeba-test.dan-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1636c09e22feac275b7124043c4e2b9ab537d1e503dfad33695f70f6dd8aabe
+size 156530
diff --git a/test/tatoeba-test.dan-spa.tsv b/test/tatoeba-test.dan-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1f4f8de872b178deec90b3cd76d35a12056daefe
--- /dev/null
+++ b/test/tatoeba-test.dan-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6247781c608b7cbce6244ec04e0d620c8c4b071f620a2b7ad4afac2bda9b522
+size 378035
diff --git a/test/tatoeba-test.dan-swe.tsv b/test/tatoeba-test.dan-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fd567a4b1d9d70ed0e45d7b1c9a8d7238eb76d43
--- /dev/null
+++ b/test/tatoeba-test.dan-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53607b21affedf513909e2af25a90e1e6dea650bba318fbf824d524cf5836eec
+size 107366
diff --git a/test/tatoeba-test.dan-tur.tsv b/test/tatoeba-test.dan-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4a91be8620849e51c68a7c2a46199a1ef004e1f1
--- /dev/null
+++ b/test/tatoeba-test.dan-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89b2fa34e0aa7a4f19007dce4409f33bbf6d8c7aa002cd89e9ddaca80f5fa7cd
+size 44770
diff --git a/test/tatoeba-test.deu-cmn_Hans.tsv b/test/tatoeba-test.deu-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5660b201ae19a02ab843d0d4739afca51d678161
--- /dev/null
+++ b/test/tatoeba-test.deu-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61e24e94772dab379c267ba2dc143f32f60a3b48c9aecfa2eacaafc249773894
+size 227663
diff --git a/test/tatoeba-test.deu-cmn_Hant.tsv b/test/tatoeba-test.deu-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..242904d81a4528a476aec4594ddddc9cf7e593eb
--- /dev/null
+++ b/test/tatoeba-test.deu-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d4eda34af5112250ddc572851a9c9488586a1250dfb12cbd6cd769c20fb8074
+size 93156
diff --git a/test/tatoeba-test.deu-deu.tsv b/test/tatoeba-test.deu-deu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dd6e19036a3c34e84ab73aa7be9178c5a2a2f9cc
--- /dev/null
+++ b/test/tatoeba-test.deu-deu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1725ceed3f0bb3bd6c900e6e3a03741b660cbaccb8cae64335b9ac9802e6b08f
+size 238013
diff --git a/test/tatoeba-test.deu-dsb.tsv b/test/tatoeba-test.deu-dsb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..435aaa85a1274abbd8d7db9ff873166fbeefcbe8
--- /dev/null
+++ b/test/tatoeba-test.deu-dsb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9750a915d3659a18a5da093a6efb8624b73e75e9efcb625c03db0efc43610b47
+size 48185
diff --git a/test/tatoeba-test.deu-ell.tsv b/test/tatoeba-test.deu-ell.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c09165cd411d33e14895c8253984cbf922196807
--- /dev/null
+++ b/test/tatoeba-test.deu-ell.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3f15908abb0ffab987f60d10fc3ed5e766e95ed0cc1bf2d0bdeb687626d8e9c
+size 238898
diff --git a/test/tatoeba-test.deu-eng.tsv b/test/tatoeba-test.deu-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bd36b1927d52d6dc76190e64cf09d855739aa62b
--- /dev/null
+++ b/test/tatoeba-test.deu-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:caf275e8d49b2e1557f7678a48bd32cb43da873ecffcb28964c6b639f6ffa7d1
+size 1595470
diff --git a/test/tatoeba-test.deu-epo.tsv b/test/tatoeba-test.deu-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ec391e77064b403bbc235d21406fe1c0b941e250
--- /dev/null
+++ b/test/tatoeba-test.deu-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d34bc53efbced35be0f6df73b1cf1a3acbc4f30f0160f81ff370fd9a38a5dfd
+size 1429646
diff --git a/test/tatoeba-test.deu-est.tsv b/test/tatoeba-test.deu-est.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..af7c74183f52d8524f7d3a162288d0e43c647d2d
--- /dev/null
+++ b/test/tatoeba-test.deu-est.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6a84f8f315e8256ec6cfe10925ec014f35d7394b6f3c0d6e703fc6d51fac05c
+size 17015
diff --git a/test/tatoeba-test.deu-eus.tsv b/test/tatoeba-test.deu-eus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c40c53a7c8609e24fc9c8225effc8f6b94e8e653
--- /dev/null
+++ b/test/tatoeba-test.deu-eus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5a03a216f0602ccfd796975149f08be3304f1855676f99a962366b16308879f
+size 35688
diff --git a/test/tatoeba-test.deu-fas.tsv b/test/tatoeba-test.deu-fas.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..089cf2b0ddd8ecb609dceeafbfd4f194296eb1ab
--- /dev/null
+++ b/test/tatoeba-test.deu-fas.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e04fb3796f2f438be3b47467f100be7925a76d74e8780c8069d6a23c6d88abc0
+size 362389
diff --git a/test/tatoeba-test.deu-fin.tsv b/test/tatoeba-test.deu-fin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f4619f84b8fcacca02825cca5c1ad0861923e96d
--- /dev/null
+++ b/test/tatoeba-test.deu-fin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:560e752fef757be7ae6b53a3e963c77cee610966d18f4f7873ba0634f41044c4
+size 207955
diff --git a/test/tatoeba-test.deu-fra.tsv b/test/tatoeba-test.deu-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6c0dca62847cd7f50b6626a6b3447ba057913a50
--- /dev/null
+++ b/test/tatoeba-test.deu-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ab27b98de7ce9f811b0ee7a63dc5073baefe9e73869e2b88bb6f34ae721bb9b
+size 1133086
diff --git a/test/tatoeba-test.deu-frr.tsv b/test/tatoeba-test.deu-frr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a697746171436671f80cd866805de411c8708feb
--- /dev/null
+++ b/test/tatoeba-test.deu-frr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd318013737b08c08fab00ee60caf065346537d2e95a732144bab8d2aa9dfa46
+size 19409
diff --git a/test/tatoeba-test.deu-gos.tsv b/test/tatoeba-test.deu-gos.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4f268cdeeb3734a1f50d67405b86303cc61808bb
--- /dev/null
+++ b/test/tatoeba-test.deu-gos.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1f864f67910e3a334153f9929ac25f7b1c3e5d51749f705aa7c32da4b38945a
+size 12260
diff --git a/test/tatoeba-test.deu-hbs.tsv b/test/tatoeba-test.deu-hbs.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6440b4de6ed4b63de0f7750791d2a3a1459dd48e
--- /dev/null
+++ b/test/tatoeba-test.deu-hbs.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cddf6c899a44b58f6db8c767aa6c82330c90aa2a0def853b46c248b1b42639a3
+size 170005
diff --git a/test/tatoeba-test.deu-heb.tsv b/test/tatoeba-test.deu-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9a89f011f7e598b5fe3d289bf404b63988cd4c00
--- /dev/null
+++ b/test/tatoeba-test.deu-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb4501060e45ccbb521c5a9c7319350ceb172479ec0a3869078b23057b93b98f
+size 305125
diff --git a/test/tatoeba-test.deu-hrv.tsv b/test/tatoeba-test.deu-hrv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..88f4bf9f70da6b4425c647e32fb22b302868f586
--- /dev/null
+++ b/test/tatoeba-test.deu-hrv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76b3b7ea47ed4493d1f8698716ef4b78b7a73dcee5c3f3357471e65dbb1bb314
+size 58300
diff --git a/test/tatoeba-test.deu-hrx.tsv b/test/tatoeba-test.deu-hrx.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7ef5f9c8110a581ab227c0bdbcea6c72119fea9a
--- /dev/null
+++ b/test/tatoeba-test.deu-hrx.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cf2546c2baea2b0c5a7839939f7dcd8b956d3cb33ff2ca14821df51088474f7
+size 32127
diff --git a/test/tatoeba-test.deu-hsb.tsv b/test/tatoeba-test.deu-hsb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5b2c5a57d8e51392ce3e4f651c2380423c857db2
--- /dev/null
+++ b/test/tatoeba-test.deu-hsb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f7b0fe0fc0ea222a2cb03c0fd8da01fa505c6f0cceed44db5a12e5e9668d0ad
+size 50779
diff --git a/test/tatoeba-test.deu-hun.tsv b/test/tatoeba-test.deu-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f61955e1ae38177c2a4593c08d934409463b5b6b
--- /dev/null
+++ b/test/tatoeba-test.deu-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c35d9c065a8faa9bdeec18ef91ed7b7d143fdfc053a9232ab7f0c7bcd16993c
+size 1351044
diff --git a/test/tatoeba-test.deu-ido.tsv b/test/tatoeba-test.deu-ido.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e41071fa6b2ae563fa2c1c7e1a0c3b7be0e4a0a2
--- /dev/null
+++ b/test/tatoeba-test.deu-ido.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c29126a93b279e5a7b119b087fa49514bb20b502c613cc03ac08780ffdd8f028
+size 58820
diff --git a/test/tatoeba-test.deu-ile.tsv b/test/tatoeba-test.deu-ile.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7c2b5c7a1837e5abf7f7f571d5e28c3ee722f643
--- /dev/null
+++ b/test/tatoeba-test.deu-ile.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:240025ed9fd4d32b6974e200f2d1b18fe2e104e557ee1f4bddd3d01cc885e57a
+size 137021
diff --git a/test/tatoeba-test.deu-ina.tsv b/test/tatoeba-test.deu-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a1346ba4c183b6280cac512706364e109ac87eeb
--- /dev/null
+++ b/test/tatoeba-test.deu-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f53bed9f8eaeb7ee7f4c8fa344d05e0a2720ddf5c03ba03b64fd59fb26b8b715
+size 137157
diff --git a/test/tatoeba-test.deu-ind.tsv b/test/tatoeba-test.deu-ind.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f859e8333dd0ad9e283c460a70bcf5ec4fc66a99
--- /dev/null
+++ b/test/tatoeba-test.deu-ind.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:151dbb4f3d38f9ce11ee81dd0cf83b746baba21b52e6639d638416f81dc231c5
+size 42298
diff --git a/test/tatoeba-test.deu-isl.tsv b/test/tatoeba-test.deu-isl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..806608e6207d627afdb4e28cd09fdf3e3d58bedc
--- /dev/null
+++ b/test/tatoeba-test.deu-isl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a44f0335dfbde64de31fca39b6c4811f370afbf371bc48242a91e3347778a8c
+size 68421
diff --git a/test/tatoeba-test.deu-ita.tsv b/test/tatoeba-test.deu-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a6c8b6cc56a6220363be48a98ed3af6890bb37fe
--- /dev/null
+++ b/test/tatoeba-test.deu-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c96283fef9fee3867c68f2d0a7e63c54a2ac15891afad55b969dc0fa5f1ea39b
+size 871314
diff --git a/test/tatoeba-test.deu-jbo.tsv b/test/tatoeba-test.deu-jbo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f59fccea7db76884bf0ad386e60624975d172d84
--- /dev/null
+++ b/test/tatoeba-test.deu-jbo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6c98159e8a5279fc863cf9a50757d121e2ad476c7af256e5ccc6da609b58d86
+size 114176
diff --git a/test/tatoeba-test.deu-jpn.tsv b/test/tatoeba-test.deu-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ea9009ca9c59baa2f946bc79b18be33b0368b1dd
--- /dev/null
+++ b/test/tatoeba-test.deu-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:978643f461773e19e801bb18f305183fb161f409eb18047f4c52b7a6b78156bc
+size 1233512
diff --git a/test/tatoeba-test.deu-jpn_Hani.tsv b/test/tatoeba-test.deu-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..667c23f71fdf4c5e34954e46db984a6b0175e0a2
--- /dev/null
+++ b/test/tatoeba-test.deu-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ed68f55727e5afe8b54465c03be5bf4f430f9aa29a32385a4929d6bd05ad851
+size 129301
diff --git a/test/tatoeba-test.deu-jpn_Hira.tsv b/test/tatoeba-test.deu-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8bf9c2537cb6247aabbbd51a45f092319d90ae33
--- /dev/null
+++ b/test/tatoeba-test.deu-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9fa36db9d4a3a98b57f6ef72b2588fd7ace99a2eaec01c6f5bf0d5e92f3b838
+size 1070014
diff --git a/test/tatoeba-test.deu-jpn_Kana.tsv b/test/tatoeba-test.deu-jpn_Kana.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..28434a9a39372caaacdbaa15316bb8a49c3abfd2
--- /dev/null
+++ b/test/tatoeba-test.deu-jpn_Kana.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b7e9d6fc99f8869275169f6f52c9636c92dc7a9bf62f608e67d59af2d50f0af
+size 33355
diff --git a/test/tatoeba-test.deu-kab.tsv b/test/tatoeba-test.deu-kab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c6bd13abc57615a2c407f9086efd43ef5fd13f7e
--- /dev/null
+++ b/test/tatoeba-test.deu-kab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:214006d8d5f3a577053f271754c4b90a3b24ab3d479a74b4deccbadf0383fd00
+size 25648
diff --git a/test/tatoeba-test.deu-kor.tsv b/test/tatoeba-test.deu-kor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..823b230c14d60c5ed4431d8489540a3ec3e6b7cb
--- /dev/null
+++ b/test/tatoeba-test.deu-kor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c687d3478f8f0648bbfa6e48344a7f66294a3ae6612e3a8b9e8e18ac4aed258
+size 110132
diff --git a/test/tatoeba-test.deu-kor_Hang.tsv b/test/tatoeba-test.deu-kor_Hang.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d84f6b75d25767ad36d0b3a454f7226cc704cac8
--- /dev/null
+++ b/test/tatoeba-test.deu-kor_Hang.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ebece1b671b37c992d15c33fab3c16d2f70f9dd97f53d63d89991634fbc7e46
+size 107534
diff --git a/test/tatoeba-test.deu-kur.tsv b/test/tatoeba-test.deu-kur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..58ad8d30782cdd8ead16c2f90c5a27119ea09934
--- /dev/null
+++ b/test/tatoeba-test.deu-kur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15f9a8ebaa42d32da6e0f770867682b93a9cc9bbaced6bbf653dadbe03159ee7
+size 16071
diff --git a/test/tatoeba-test.deu-kur_Latn.tsv b/test/tatoeba-test.deu-kur_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4c1604d0ec2abe184885c8141f8274ac7bcd05b2
--- /dev/null
+++ b/test/tatoeba-test.deu-kur_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d62e3a5ff453b6a6bd22ddb3e2bd98916c108ff0b5ef576bc6be85f65d2e5024
+size 14913
diff --git a/test/tatoeba-test.deu-lad.tsv b/test/tatoeba-test.deu-lad.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5d7cdb02ba4a8f7dd169853f0c3e51ba3b74da57
--- /dev/null
+++ b/test/tatoeba-test.deu-lad.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66f78e4a826dccb9e8b202ba49769db23648ee4b13e3eaf6bc8cb0f524bf0f41
+size 14517
diff --git a/test/tatoeba-test.deu-lat.tsv b/test/tatoeba-test.deu-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..129dd675052ea68a1a0042ed0818a1ee097243ed
--- /dev/null
+++ b/test/tatoeba-test.deu-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccb38071ec8c9ee15ce6e08f860294d3241d61a54d0896ea2155ea40522313b8
+size 142634
diff --git a/test/tatoeba-test.deu-lfn.tsv b/test/tatoeba-test.deu-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..93c6af5120e96e9b3df67944a7aba860c4d4b4cd
--- /dev/null
+++ b/test/tatoeba-test.deu-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd9293516047dbc70c1d55006787af15136542f7d178fdcf3fe321ce6ba1a7c3
+size 43327
diff --git a/test/tatoeba-test.deu-lfn_Latn.tsv b/test/tatoeba-test.deu-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5a884e2ca89eade927db2a36f32df3b9ef863c09
--- /dev/null
+++ b/test/tatoeba-test.deu-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b50336bd12f3146a352cb340b73b6a45327daba8216a7669f141377ea27abbd
+size 25836
diff --git a/test/tatoeba-test.deu-lit.tsv b/test/tatoeba-test.deu-lit.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d3825634add9c198ab5c7909f9e314d10d27b79e
--- /dev/null
+++ b/test/tatoeba-test.deu-lit.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c286294065073c0e44d7db45ccf7f65985d0801fdb790dbb5339c23ac9c68010
+size 92530
diff --git a/test/tatoeba-test.deu-ltz.tsv b/test/tatoeba-test.deu-ltz.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..aaa817be950f1b857be3bff83057701574db117a
--- /dev/null
+++ b/test/tatoeba-test.deu-ltz.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f929569ee4191b883839c62336a7ec0f09ce134d76976d43b49ee1595882a8e8
+size 23796
diff --git a/test/tatoeba-test.deu-msa.tsv b/test/tatoeba-test.deu-msa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0aeabfd8e993be4ca7a171aae94ca60bd9410086
--- /dev/null
+++ b/test/tatoeba-test.deu-msa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4c221a6a73713238db5357d12f0b5b36ec2d4582c7ff708a083a201437396c5
+size 46985
diff --git a/test/tatoeba-test.deu-nds.tsv b/test/tatoeba-test.deu-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e560d27ee08d3354cd30712d780fde7335c0bfc7
--- /dev/null
+++ b/test/tatoeba-test.deu-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:241b3252c1dad2301e79e4dc31b0522924428b530e303add358a17b31e6ae927
+size 796333
diff --git a/test/tatoeba-test.deu-nld.tsv b/test/tatoeba-test.deu-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..486c84034b4a187336be8e4ab94a432fad40fc8c
--- /dev/null
+++ b/test/tatoeba-test.deu-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:641198e48e2be93d81dcdffdce6953987a669661081b07161dd98bea000f2bf2
+size 811815
diff --git a/test/tatoeba-test.deu-nob.tsv b/test/tatoeba-test.deu-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e36979a1873ae4286da123f56ce9d15e5aacbb9d
--- /dev/null
+++ b/test/tatoeba-test.deu-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a22ae0766ada461f7b1f046b152dd7c53af146d2b6b4f175348e3ef4eda6ab23
+size 363529
diff --git a/test/tatoeba-test.deu-nor.tsv b/test/tatoeba-test.deu-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b5516d49f87a3b62ca93760e16e0d8e790493f62
--- /dev/null
+++ b/test/tatoeba-test.deu-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:116c496df84848b24160398bdca852395fb2f64c13aa55444f170f29449c2087
+size 373631
diff --git a/test/tatoeba-test.deu-pol.tsv b/test/tatoeba-test.deu-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..edf743a3d438c48fae811a6bc50ee3a97db6df35
--- /dev/null
+++ b/test/tatoeba-test.deu-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39c4b4485ffb3e79f72b4ce8655ca30f54708e47f9dd4f57b2a76085dc446ec9
+size 405447
diff --git a/test/tatoeba-test.deu-por.tsv b/test/tatoeba-test.deu-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..59463dff8249866b5a1e89b20718247c98be4ac2
--- /dev/null
+++ b/test/tatoeba-test.deu-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd08e63b3440f366264f68fdf195f9b87ddd5ae30a0597057a49499f7f0d694a
+size 894266
diff --git a/test/tatoeba-test.deu-ron.tsv b/test/tatoeba-test.deu-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fa708b97c4f344230d00c6099bc83ae25f0e0e57
--- /dev/null
+++ b/test/tatoeba-test.deu-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed50c25914fb485f92d8b835efc1580ea292cd571724fdb50efb68861e64e529
+size 84604
diff --git a/test/tatoeba-test.deu-run.tsv b/test/tatoeba-test.deu-run.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ce3131a71bcdc20da13a8f90e01f1d5906bf708c
--- /dev/null
+++ b/test/tatoeba-test.deu-run.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a20725afd107c866b73b548086cffb6c669b56b75e89d9d3f3f78aca564e332
+size 100540
diff --git a/test/tatoeba-test.deu-rus.tsv b/test/tatoeba-test.deu-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b73eae7115062297cabd27d989173bdd0786535e
--- /dev/null
+++ b/test/tatoeba-test.deu-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e01735e401a2f363a4b4306bcaff9a11d4281a79bf52a6f7a916a567e76ebac2
+size 1345955
diff --git a/test/tatoeba-test.deu-slv.tsv b/test/tatoeba-test.deu-slv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..374c69e35792260480898065942f390a0fb7d5c6
--- /dev/null
+++ b/test/tatoeba-test.deu-slv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98402d4f1a6ad1d63dddc416f7582d8b60420815acd781e918f9755f2dd4b846
+size 30184
diff --git a/test/tatoeba-test.deu-spa.tsv b/test/tatoeba-test.deu-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..06647062e1b1bd4cf0e6b29116a8764cbf870a1c
--- /dev/null
+++ b/test/tatoeba-test.deu-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a5ab64548dd6d98dc29b9a2732a8bbf4aaa729f0fc22677c6c6286ad3cbbd66
+size 930791
diff --git a/test/tatoeba-test.deu-srp_Latn.tsv b/test/tatoeba-test.deu-srp_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..73cd59eb5bb7414ed4bacb1e7452c6870476486d
--- /dev/null
+++ b/test/tatoeba-test.deu-srp_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d459947eff187ba3007684fe5da74c641012c5f2b8683d9b05d1eb871e2e2cce
+size 94067
diff --git a/test/tatoeba-test.deu-swe.tsv b/test/tatoeba-test.deu-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4acc12a59f1ea18f08ff81c60736b61343f97fe4
--- /dev/null
+++ b/test/tatoeba-test.deu-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46535654240d4837c30949b807324cde9b68d262f39f4b9c4ddc5c95ed56d399
+size 254512
diff --git a/test/tatoeba-test.deu-swg.tsv b/test/tatoeba-test.deu-swg.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f85b56bddeeb745f85915a44c018f867e4b6b12a
--- /dev/null
+++ b/test/tatoeba-test.deu-swg.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e500595a843e17cd52c1b31460ad06361e33eff089c8c5676c32da47b60ff34f
+size 166882
diff --git a/test/tatoeba-test.deu-tat.tsv b/test/tatoeba-test.deu-tat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..265f3e06e7ee07b3339750d681e0e6e4c3e0ca9a
--- /dev/null
+++ b/test/tatoeba-test.deu-tat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26fa6b36874e72cfd8077c2055dc741174c2977df41c9a580df8b65b6902e2af
+size 371833
diff --git a/test/tatoeba-test.deu-tgl.tsv b/test/tatoeba-test.deu-tgl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5f754d87df89a4bbf1441bdb395b2c9b2f4578c0
--- /dev/null
+++ b/test/tatoeba-test.deu-tgl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:427d2b2579c9fcabab268e64fb5ada2ccf12722d12cdb7ab027e35d8f03a5a80
+size 28456
diff --git a/test/tatoeba-test.deu-tlh.tsv b/test/tatoeba-test.deu-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..45f1ac0d6fdda3ed9e6436d133bf21e9d3d5d39c
--- /dev/null
+++ b/test/tatoeba-test.deu-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5e9afa800f542e419f9f2c9264318f0c662bf7c1613714153ad83e572f279da
+size 78188
diff --git a/test/tatoeba-test.deu-toki.tsv b/test/tatoeba-test.deu-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..36342a8241f94fb68db4e6e5f38bd2ac90d4384d
--- /dev/null
+++ b/test/tatoeba-test.deu-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac664313e0505c71e636248c25415095296e8bd0b7bfe18cf672cd88e6e6c7b2
+size 945160
diff --git a/test/tatoeba-test.deu-tur.tsv b/test/tatoeba-test.deu-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7260786ec6bb3d02beb0358b8d651a8f341af729
--- /dev/null
+++ b/test/tatoeba-test.deu-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f2f768cc9433b9166d94ff361008dc657cb9ed2d553374f1713acf92e597e43
+size 409984
diff --git a/test/tatoeba-test.deu-ukr.tsv b/test/tatoeba-test.deu-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2ba621639dfd81eab071f8c6a7c27ebdf8f507f1
--- /dev/null
+++ b/test/tatoeba-test.deu-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:296afb268d70a498057f99def4b6703895b9b7770741c1f7f2c72fcd520c0197
+size 864322
diff --git a/test/tatoeba-test.deu-vie.tsv b/test/tatoeba-test.deu-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..599b5450e69165251892f0b2d8bef2728f2b49c1
--- /dev/null
+++ b/test/tatoeba-test.deu-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da912aef04066422d8e66acf6492ad8d5d45999ec5902cbee7345a59b7a5cce1
+size 38820
diff --git a/test/tatoeba-test.deu-vol.tsv b/test/tatoeba-test.deu-vol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7a1b68b755af71983a6f2b9439245afbdf811304
--- /dev/null
+++ b/test/tatoeba-test.deu-vol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff488e1642bce3ec7df8747a5ea72368720c330284629fe6fa643c50b4bc0543
+size 14006
diff --git a/test/tatoeba-test.deu-yid.tsv b/test/tatoeba-test.deu-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2a9e6dc447d3db61e101237862e93dfe242989ed
--- /dev/null
+++ b/test/tatoeba-test.deu-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90062efa03fff1c5f62899a906338993eccaa70c0e0e0528056121436def44cf
+size 79442
diff --git a/test/tatoeba-test.deu-zho.tsv b/test/tatoeba-test.deu-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c4c555375f18223107a62aa4c0936bb952a15e59
--- /dev/null
+++ b/test/tatoeba-test.deu-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78c698166b4e143caefd5a34291ff621c893add135f8cad2cdcb72fa454675f3
+size 327326
diff --git a/test/tatoeba-test.dsb-hsb.tsv b/test/tatoeba-test.dsb-hsb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5fc3a27baf1b30aed5a88b217ca20c091f298d1d
--- /dev/null
+++ b/test/tatoeba-test.dsb-hsb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53a39905982ed232d73eeff5717ddc19493173073c6eb9c803f16dd595acffb1
+size 15387
diff --git a/test/tatoeba-test.dsb-slv.tsv b/test/tatoeba-test.dsb-slv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6772b0477b35edfc2a912ac2757b5f700c49170f
--- /dev/null
+++ b/test/tatoeba-test.dsb-slv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec2cf1b60bbe2d93cb7b721bcacde8d05cd5113602b29b871ddaa9f0eff6b0d5
+size 16248
diff --git a/test/tatoeba-test.dtp-eng.tsv b/test/tatoeba-test.dtp-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f7e33cb604d8eaa39f0c5c7ea5426f4b6f25b4b6
--- /dev/null
+++ b/test/tatoeba-test.dtp-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67bac3a28525771bd487cbc35f264201ca87203e2989a3b978e3107a680e8e94
+size 141337
diff --git a/test/tatoeba-test.dtp-jpn.tsv b/test/tatoeba-test.dtp-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..57fe0adafa8e5a9f201a79ac2d6ef5d45dd697a2
--- /dev/null
+++ b/test/tatoeba-test.dtp-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c967c6efead7a6c0f3197697e22df592d6b35158ca60dca9a89e576533b94484
+size 21300
diff --git a/test/tatoeba-test.dtp-jpn_Hira.tsv b/test/tatoeba-test.dtp-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ef5d77b823d2679f0c0f9a84f21d884f07d0ce0d
--- /dev/null
+++ b/test/tatoeba-test.dtp-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b54d8b168c00ccc48706925513a10a1cbebea1a1c620c468cd56a95868e7844
+size 20048
diff --git a/test/tatoeba-test.dtp-msa.tsv b/test/tatoeba-test.dtp-msa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..89c4101ed8d5ae52cd0f30b8fa9e1ddc52f40d05
--- /dev/null
+++ b/test/tatoeba-test.dtp-msa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d165ffd5c9b2fc6929585ef1b26d0f65ad328c55599301580f8ad73ef751492
+size 53688
diff --git a/test/tatoeba-test.dtp-zsm_Latn.tsv b/test/tatoeba-test.dtp-zsm_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d315706cf96fb9d83d3f9268548114d4548b2e2f
--- /dev/null
+++ b/test/tatoeba-test.dtp-zsm_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60861f12102d0626b7ae8e052a7ff04894262b3164201316845702b4f56bd9ef
+size 52979
diff --git a/test/tatoeba-test.egl-ita.tsv b/test/tatoeba-test.egl-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..26bbe6d196ed76c0e545c8f752a7296bfa741028
--- /dev/null
+++ b/test/tatoeba-test.egl-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da608ae0df250e515476e6d7507d312be96ee608d0e229d52e9e20b8186a6843
+size 10674
diff --git a/test/tatoeba-test.ell-ell.tsv b/test/tatoeba-test.ell-ell.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6f89feee8f27fbd0cb53304fe91b530ab054465b
--- /dev/null
+++ b/test/tatoeba-test.ell-ell.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65e24d86f98516d87dbb07973c788f6cfea7ecd0624072e9f1efe5c2f20f33fa
+size 57805
diff --git a/test/tatoeba-test.ell-eng.tsv b/test/tatoeba-test.ell-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c5c8b82c371138a0b3a2ae464e1e2d76f4048a9f
--- /dev/null
+++ b/test/tatoeba-test.ell-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb29a27124d06b3f1b015ee67fc78f7918838136bb74d9e645f830ac460aa0fa
+size 963409
diff --git a/test/tatoeba-test.ell-epo.tsv b/test/tatoeba-test.ell-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c0ed399d6c3e23c23a5e3d6c28d14f24ec82f2e1
--- /dev/null
+++ b/test/tatoeba-test.ell-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a796cf20814fb59fe2e944955723508b8bfcd3222a0cae3a21364bf82f48edb
+size 58730
diff --git a/test/tatoeba-test.ell-fra.tsv b/test/tatoeba-test.ell-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..65488930d46172a9c528916c024a0a0ffe7c11cc
--- /dev/null
+++ b/test/tatoeba-test.ell-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41dbb14d8e7a8e9620c9c95b07c37bd5c3c794bd3cdd754e651b57b5951e6ec5
+size 135841
diff --git a/test/tatoeba-test.ell-ita.tsv b/test/tatoeba-test.ell-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b73486ef57974d58d3eb8e78d9caf71c31453204
--- /dev/null
+++ b/test/tatoeba-test.ell-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ea54cefa2cc103814be380e0a420f0475059037faf1d81ea77c46535be0317e
+size 36093
diff --git a/test/tatoeba-test.ell-nld.tsv b/test/tatoeba-test.ell-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7c9eb2007ae814824b7c241ffb08f6177c025ca3
--- /dev/null
+++ b/test/tatoeba-test.ell-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79c3e19b547f6fa16e4b59e349b711dbd051bf6df5f3ab8fe7397ec9e3452fe5
+size 121339
diff --git a/test/tatoeba-test.ell-por.tsv b/test/tatoeba-test.ell-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6e93488ccbf08c96b24258928ddf4ebe530fcd18
--- /dev/null
+++ b/test/tatoeba-test.ell-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0512cf3d40ccd990556a0e4bc8a84787956df8b8d3d2917e8e3341e2ee8110d2
+size 72136
diff --git a/test/tatoeba-test.ell-rus.tsv b/test/tatoeba-test.ell-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e34440009ed8ad43c398dafd2811a528b957a73f
--- /dev/null
+++ b/test/tatoeba-test.ell-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06ae30405a86948245f822206c6bd0965bb2d3ecf9fea3a68daaf4b3811e0ef6
+size 282508
diff --git a/test/tatoeba-test.ell-spa.tsv b/test/tatoeba-test.ell-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2b93ade77028372c9e4f32ec5c3d76a4f4ef7090
--- /dev/null
+++ b/test/tatoeba-test.ell-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c34c78933aaf7383b9fd9d941e817fee25324f3f08ff7b28bceceb6054bbe40
+size 155748
diff --git a/test/tatoeba-test.ell-swe.tsv b/test/tatoeba-test.ell-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fdbfa25633299a67a560617e198dfeb9f5647cf8
--- /dev/null
+++ b/test/tatoeba-test.ell-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:840c9cf19e6c93d9955ec653dd34fdf77248a5672a04966d3847a6db1a2a4d45
+size 21696
diff --git a/test/tatoeba-test.ell-tur.tsv b/test/tatoeba-test.ell-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2d5cbcbefa32039afebe28924cd82501fad5ce8d
--- /dev/null
+++ b/test/tatoeba-test.ell-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:919cc77fc010329e1d8e30060114368fbb238a3c362ae51f6bf366b9a5bbc4f3
+size 132670
diff --git a/test/tatoeba-test.eng-bos_Latn.tsv b/test/tatoeba-test.eng-bos_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..548a8311d540d4489faab89a9e37e5f871313463
--- /dev/null
+++ b/test/tatoeba-test.eng-bos_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58b1607d294301f61c590a864c6928f9361f8b2397a3b28e50d049c2c2aa6e03
+size 19082
diff --git a/test/tatoeba-test.eng-cmn_Hans.tsv b/test/tatoeba-test.eng-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dbd12ee84e56be0e61ab6a468b44ba6b68650ecb
--- /dev/null
+++ b/test/tatoeba-test.eng-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a756bfa75046ad24dcda07e3bd49352f48a718d49b649ac00b85fa2d833b0b1
+size 385634
diff --git a/test/tatoeba-test.eng-cmn_Hant.tsv b/test/tatoeba-test.eng-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ea824ee6c80d889abd3af05167c7d2cc8a4a4bd3
--- /dev/null
+++ b/test/tatoeba-test.eng-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:399e9c7d5a4ccbcbea2c737439deda1480ef64c225f796f79eb9152f608d132b
+size 360736
diff --git a/test/tatoeba-test.eng-eng.tsv b/test/tatoeba-test.eng-eng.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4741fd79b3c047a3394114286749ec6e463b34fb
--- /dev/null
+++ b/test/tatoeba-test.eng-eng.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80d27e891ea94b003fc81e0964ed4924956db233d667b67243a1670e3d685d46
+size 1124615
diff --git a/test/tatoeba-test.eng-epo.tsv b/test/tatoeba-test.eng-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..687452c7ee1ab508df9691bf2783e1f405ccf93d
--- /dev/null
+++ b/test/tatoeba-test.eng-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3551bc6b128c0900dda530ba28c1e5e3555a4ec60b0be4f8d335569ce0e2859f
+size 1455517
diff --git a/test/tatoeba-test.eng-est.tsv b/test/tatoeba-test.eng-est.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cafe4092ee7d09a36f1dcc87da72563088c68881
--- /dev/null
+++ b/test/tatoeba-test.eng-est.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e34a640d36387c9d1dd5ecae7bbdf8b103bd062f278a91cbbf8830338a9d87a1
+size 87767
diff --git a/test/tatoeba-test.eng-eus.tsv b/test/tatoeba-test.eng-eus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3338bed8662a342f9abf17ac61b3fbd46a02b965
--- /dev/null
+++ b/test/tatoeba-test.eng-eus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdb6dc2e7dcc042a92488f38c5d9adce4d813be985b3c722ed9810d2c1fb0606
+size 83906
diff --git a/test/tatoeba-test.eng-fao.tsv b/test/tatoeba-test.eng-fao.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9b9fc9c4ad2b736e7ca6220bcb333f7b0d3ed671
--- /dev/null
+++ b/test/tatoeba-test.eng-fao.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f15986dfa0d0fabe9fe241ab46e4721cee0fbc175fe3747c7384614e0264b297
+size 20156
diff --git a/test/tatoeba-test.eng-fas.tsv b/test/tatoeba-test.eng-fas.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1a73a7c17327c50febba09f11a532ed9168f71fc
--- /dev/null
+++ b/test/tatoeba-test.eng-fas.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ec475e7d72eaeccfa013967e952ec8d0533fa1caff543c8ba50524ccc6abc0d
+size 410415
diff --git a/test/tatoeba-test.eng-fin.tsv b/test/tatoeba-test.eng-fin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bb5b34ec378325d5300f3a1db616bbd53edc3167
--- /dev/null
+++ b/test/tatoeba-test.eng-fin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:198558e2c1b9b61ffc9e8a5618f17c638ddef78ddffc105bb0138586dd282164
+size 843346
diff --git a/test/tatoeba-test.eng-fra.tsv b/test/tatoeba-test.eng-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..184f1101fce3774def39063c0ea3005ae6dce7a1
--- /dev/null
+++ b/test/tatoeba-test.eng-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b052c30b94c4ed294abae539d7d75e431c84ba43a663e34dbc9e8d2bf8d5efbb
+size 1106067
diff --git a/test/tatoeba-test.eng-fry.tsv b/test/tatoeba-test.eng-fry.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bdd3f39e6e289a9a07b2805131b8127526524b2d
--- /dev/null
+++ b/test/tatoeba-test.eng-fry.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:412c422f21c67e3ef33f5c96b8c19657c2691f36a961f0d9bc937ae9dd822902
+size 15737
diff --git a/test/tatoeba-test.eng-gla.tsv b/test/tatoeba-test.eng-gla.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6780894e8722be0564c1a2ad2bdc25560bb0dde0
--- /dev/null
+++ b/test/tatoeba-test.eng-gla.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66a400eb7458171367dbff716f60e4da6f921e3cbc658776da2cfcca5c8b9084
+size 68841
diff --git a/test/tatoeba-test.eng-gle.tsv b/test/tatoeba-test.eng-gle.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7f20f9754f7df46945829516e375082e67ac0a5b
--- /dev/null
+++ b/test/tatoeba-test.eng-gle.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:633f5ca499570e5972ace0988af4197ae9f41ccd3b4098577c698b3f4d53dd6d
+size 118507
diff --git a/test/tatoeba-test.eng-glg.tsv b/test/tatoeba-test.eng-glg.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..07175d89014f124d72bc7051c658b972fc477621
--- /dev/null
+++ b/test/tatoeba-test.eng-glg.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:788599047b0c9e674a4527b6265eda166643ca9581c863cf48fd2cb1ba82eca1
+size 85016
diff --git a/test/tatoeba-test.eng-gos.tsv b/test/tatoeba-test.eng-gos.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f7c660b5fe3078e4c502472c30eecb1deefe3da5
--- /dev/null
+++ b/test/tatoeba-test.eng-gos.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9da0e300d56ab561bc5bffe46ee3fc10b329cb0dd91f8d011f32795a820314af
+size 56438
diff --git a/test/tatoeba-test.eng-got.tsv b/test/tatoeba-test.eng-got.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f659da3961f33eba7616a81c03a87be71bdc519c
--- /dev/null
+++ b/test/tatoeba-test.eng-got.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39abc7e86aa8bad023866a17b29b96ae6c348a3f5084f5f8156184ab224d8023
+size 20327
diff --git a/test/tatoeba-test.eng-grc.tsv b/test/tatoeba-test.eng-grc.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..228eb7f899d06ae30fe6e101fd401d09d9af1d3a
--- /dev/null
+++ b/test/tatoeba-test.eng-grc.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e85de09abbdd4f3f84f8d53c67cffefc7113d48a5e9778016b6b07d8e9a2b15
+size 70785
diff --git a/test/tatoeba-test.eng-gsw.tsv b/test/tatoeba-test.eng-gsw.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..edbf5b4d6fcdda84992f569e4e3b6063b90ca02b
--- /dev/null
+++ b/test/tatoeba-test.eng-gsw.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f9b3aef1efeaf78037e652f1f6e8f1cecbd6ae076e3284290dfded4a70aa07b
+size 9985
diff --git a/test/tatoeba-test.eng-hbs.tsv b/test/tatoeba-test.eng-hbs.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2e8c4371c59ee1d337c687d5b4581e8a57896644
--- /dev/null
+++ b/test/tatoeba-test.eng-hbs.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3c4618fd7f49b5ab92094270cc2e6c93655b741cd4fa025ee0ecdcc04c063ed
+size 757053
diff --git a/test/tatoeba-test.eng-heb.tsv b/test/tatoeba-test.eng-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3b3876e6019a5a94937c33b2256f31ff9c02ec5f
--- /dev/null
+++ b/test/tatoeba-test.eng-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:482f97e0bcc3516f2a5404dbd04e7aaf3539c2f11f6a75952b122cb51e10e3af
+size 903147
diff --git a/test/tatoeba-test.eng-hin.tsv b/test/tatoeba-test.eng-hin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6fa0927fb2ab4e4a667b8dea96bb0c1afc571824
--- /dev/null
+++ b/test/tatoeba-test.eng-hin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd3b0f47683e2fa0852e2203da84952b9f088f578702f49a8d81c88261b81c42
+size 572488
diff --git a/test/tatoeba-test.eng-hoc.tsv b/test/tatoeba-test.eng-hoc.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5818285593f52b3a0281b6d0372304ba644e798e
--- /dev/null
+++ b/test/tatoeba-test.eng-hoc.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4684552feb4010258610ab6f4cba66b87fe90b9e832f4a9d704eaad88134383b
+size 36715
diff --git a/test/tatoeba-test.eng-hoc_Latn.tsv b/test/tatoeba-test.eng-hoc_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b52eea578ad8c71a9165cc387e5d36ee01ef7a00
--- /dev/null
+++ b/test/tatoeba-test.eng-hoc_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30b6cb229b2ff4d24133d7ac5c64137d337fda82d4b4f1a5668f2ba1102940f5
+size 36698
diff --git a/test/tatoeba-test.eng-hrv.tsv b/test/tatoeba-test.eng-hrv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3721dac2f3f82dbb01ab1d9ab8953b16d47b55cd
--- /dev/null
+++ b/test/tatoeba-test.eng-hrv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a344c17d7815940a2440086024fd95d0b56c4d2f0c23453fde092332cbb90319
+size 104422
diff --git a/test/tatoeba-test.eng-hrx.tsv b/test/tatoeba-test.eng-hrx.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..043dc44b5b8d974c4eb04dd97f6ad71040d91107
--- /dev/null
+++ b/test/tatoeba-test.eng-hrx.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d205e3a024711e5e4b815c73ca9ddd1e097b2154f0be5ca4a2a45d6cdda6f33
+size 13943
diff --git a/test/tatoeba-test.eng-hun.tsv b/test/tatoeba-test.eng-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..80947afcd973d3ffffb499427f8c36568ae3dc59
--- /dev/null
+++ b/test/tatoeba-test.eng-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f73e7d8e58c827d5a624c8a31b4684ca970ab6d8e6a6719d2c15f3d3d5b543b
+size 959274
diff --git a/test/tatoeba-test.eng-hye.tsv b/test/tatoeba-test.eng-hye.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6c54ad2f4fb29229bc5c36428e06f6a06b766f4c
--- /dev/null
+++ b/test/tatoeba-test.eng-hye.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7240fc77e45002ac446039819b29a663845421f8f484d6c03119503ffa28d0a1
+size 73369
diff --git a/test/tatoeba-test.eng-ido.tsv b/test/tatoeba-test.eng-ido.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fd4b4153f92f111650a27f2867b6e428b6285603
--- /dev/null
+++ b/test/tatoeba-test.eng-ido.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8678087a2b7c949df51896f2a9c6b126813ad9bceaaf3176e438f26f2a62992b
+size 149129
diff --git a/test/tatoeba-test.eng-ido_Latn.tsv b/test/tatoeba-test.eng-ido_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4e3a50ae6d42e3f0009ec67a4d5f7254a1793704
--- /dev/null
+++ b/test/tatoeba-test.eng-ido_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8eff68099bead1878f783fdc9c1ea740fea9a14d2270f2100819866b7ddb94dc
+size 149043
diff --git a/test/tatoeba-test.eng-ile.tsv b/test/tatoeba-test.eng-ile.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..98cc2d07a163fc77bb30fb90d2d1c002788a70a5
--- /dev/null
+++ b/test/tatoeba-test.eng-ile.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fb1642836b3e9f7e53789b688ff286720fad7344b758fae5a4227257c04ecd9
+size 112552
diff --git a/test/tatoeba-test.eng-ilo.tsv b/test/tatoeba-test.eng-ilo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..edd262852b43d45f2348b3cb781aa6756c667d55
--- /dev/null
+++ b/test/tatoeba-test.eng-ilo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59168bab38f6baaa0e76815b6e133144172b438731c57658671d3e9843942e9d
+size 78404
diff --git a/test/tatoeba-test.eng-ina.tsv b/test/tatoeba-test.eng-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..407a4c6b1fcdf4f03799b2967c853b4bf66b8014
--- /dev/null
+++ b/test/tatoeba-test.eng-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fe8dc3e704343e0693d5d405bc72f7b75cf298627372f877ac75d2c53979e19
+size 473922
diff --git a/test/tatoeba-test.eng-ind.tsv b/test/tatoeba-test.eng-ind.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fb373766b614f5d984e004f23524b4375c14fe52
--- /dev/null
+++ b/test/tatoeba-test.eng-ind.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7af9c45b3bd246351bfb065d3f2a9e1ed1834771f87f02fccb3b61dd576989c5
+size 329610
diff --git a/test/tatoeba-test.eng-isl.tsv b/test/tatoeba-test.eng-isl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e152070b1b271779b5282f3e5bbd9817d3750219
--- /dev/null
+++ b/test/tatoeba-test.eng-isl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:657d23bcd3456c6f1c8118a95e450e97932a7f525c9abe8855c8eac5dc29586d
+size 210952
diff --git a/test/tatoeba-test.eng-ita.tsv b/test/tatoeba-test.eng-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ec5f0a9a3016b5cf878fec10fbae9505c5944962
--- /dev/null
+++ b/test/tatoeba-test.eng-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9453277ab8b9f1d9b2d4f4839060ec83b66811b539c2b604a44ff9f8f188cb95
+size 1233733
diff --git a/test/tatoeba-test.eng-jav.tsv b/test/tatoeba-test.eng-jav.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ab7950b8f59a431fd25dffb41fc593d61689bc84
--- /dev/null
+++ b/test/tatoeba-test.eng-jav.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:feb7cd6cff36abe6d9451bf789a75e351b51d553c9a511aabc463d84c169b634
+size 19245
diff --git a/test/tatoeba-test.eng-jbo.tsv b/test/tatoeba-test.eng-jbo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..45baee7f60c3136e455c1e75c07ab0a8c4ffcb47
--- /dev/null
+++ b/test/tatoeba-test.eng-jbo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6a5e467341e740684609edfa3112cc6dd8334aab28ecae19863ca87a418e199
+size 364043
diff --git a/test/tatoeba-test.eng-jbo_Latn.tsv b/test/tatoeba-test.eng-jbo_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..50f3eedd19d07e58ea9130ad220ae9d3ab5e1024
--- /dev/null
+++ b/test/tatoeba-test.eng-jbo_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c72c84e3f216f23d3bcf2b4f5cb44594bbee00b30e7992cee1b4784142655a61
+size 363806
diff --git a/test/tatoeba-test.eng-jpn.tsv b/test/tatoeba-test.eng-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d9138a08455ead5c845043e8fc1a5e9f5d7ef3d4
--- /dev/null
+++ b/test/tatoeba-test.eng-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a6b15982bd3472582d3ffb753d984196d9d433d45d0daf993f04786ac053a32
+size 1491702
diff --git a/test/tatoeba-test.eng-jpn_Hani.tsv b/test/tatoeba-test.eng-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..07b0a141a1a98220b5d35ddfec92d4e0b640d838
--- /dev/null
+++ b/test/tatoeba-test.eng-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09dc88e1fdde2cd4f60bbeebe655d77cb67f26b05307532e131e6f7537d68b8e
+size 147492
diff --git a/test/tatoeba-test.eng-jpn_Hira.tsv b/test/tatoeba-test.eng-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3f58a377bb00e253145ed5db98f62e99c874b5b5
--- /dev/null
+++ b/test/tatoeba-test.eng-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e73623b1ff9881d912da2ac8d9b5468f476e9bbdb31baedbca5511ad62159d4a
+size 1289993
diff --git a/test/tatoeba-test.eng-jpn_Kana.tsv b/test/tatoeba-test.eng-jpn_Kana.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..30b63e91a3688ed947d90bf5c3a72ac34892305c
--- /dev/null
+++ b/test/tatoeba-test.eng-jpn_Kana.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e5e3dd4eb29c07e38de17486ac4f0ebce7a198f0fd3d38be54fb90f308e25e4
+size 51632
diff --git a/test/tatoeba-test.eng-kab.tsv b/test/tatoeba-test.eng-kab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ac30580a65353bb442e3f8d53f7c9e426632d077
--- /dev/null
+++ b/test/tatoeba-test.eng-kab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3710553fadfca2a5135ff2da6c8d65b417180fc5c2bcdaee01d9d6922c1af1c0
+size 812807
diff --git a/test/tatoeba-test.eng-kat.tsv b/test/tatoeba-test.eng-kat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..352e1b6deea498642caf285f1cee64b1808f8908
--- /dev/null
+++ b/test/tatoeba-test.eng-kat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af45cebe895d812e425e42807c9ed7a1896ac513ca00e5e912154d81872b6469
+size 94570
diff --git a/test/tatoeba-test.eng-kaz.tsv b/test/tatoeba-test.eng-kaz.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0c02bf1710b1987005c3abf387185cc776149dd2
--- /dev/null
+++ b/test/tatoeba-test.eng-kaz.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41edcb5bf13ae3a07a4cf1cb98a400280ed635fcbd5989585a7c4b4a5b53773d
+size 37502
diff --git a/test/tatoeba-test.eng-kaz_Cyrl.tsv b/test/tatoeba-test.eng-kaz_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8d3f6a38bb76887c26fedc58d231df75ab82c97d
--- /dev/null
+++ b/test/tatoeba-test.eng-kaz_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b34a7026c1b1abae47ac510bd6ad0c1d263dbd47275feeaae7df262175fabcf
+size 36935
diff --git a/test/tatoeba-test.eng-kha.tsv b/test/tatoeba-test.eng-kha.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..495a06b4e70815e1a478253b490c90c3eeae0ee2
--- /dev/null
+++ b/test/tatoeba-test.eng-kha.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7dfa79395c841b4ffa00b9cddf10f41aa01011c187f9d961acd403aba4d1517e
+size 84084
diff --git a/test/tatoeba-test.eng-khm.tsv b/test/tatoeba-test.eng-khm.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2bcc2fe0921cb20e5a5a75d3be9e10dd342ccc3f
--- /dev/null
+++ b/test/tatoeba-test.eng-khm.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2c238755471923923e7fb04b674deaf85158938c0c30b0ffb7ec2d5e51b0444
+size 71191
diff --git a/test/tatoeba-test.eng-kor.tsv b/test/tatoeba-test.eng-kor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7158aa6e6039a9201b76ed28250ae6326c7bb0fb
--- /dev/null
+++ b/test/tatoeba-test.eng-kor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:616ea9978726a281eed41e2edca1ccb82776a38f58fbc44eab8ecedf2afe138e
+size 208033
diff --git a/test/tatoeba-test.eng-kor_Hang.tsv b/test/tatoeba-test.eng-kor_Hang.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cb4ab580e18a49c263459c2e51234d0fb4e4c32c
--- /dev/null
+++ b/test/tatoeba-test.eng-kor_Hang.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b477083d41ce5bc862699cc948aa701474154ff89ab376eb6ddfaaffaefc5dd6
+size 206579
diff --git a/test/tatoeba-test.eng-kur.tsv b/test/tatoeba-test.eng-kur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..36966eb207b2840039ba039e04247119b2e77ffc
--- /dev/null
+++ b/test/tatoeba-test.eng-kur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9ca36534e21ea0a94a2b3471feeb9f25610a0ec857aa668ec6ca8b08f387c3f
+size 30895
diff --git a/test/tatoeba-test.eng-kur_Latn.tsv b/test/tatoeba-test.eng-kur_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2a6ba23f3c0df73b97810db89d2175832914314e
--- /dev/null
+++ b/test/tatoeba-test.eng-kur_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76a3854ff9d93a0d29f34cb69a65a0d6c1a9bb44f89871defb3ad799c061c71e
+size 18697
diff --git a/test/tatoeba-test.eng-kzj.tsv b/test/tatoeba-test.eng-kzj.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..52512fa79e246f586e4c34408b92773436cea38a
--- /dev/null
+++ b/test/tatoeba-test.eng-kzj.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:597e815e819b2081ea7ca796699e1e7afa4e34d3285dae78ef461e9781c405e0
+size 107218
diff --git a/test/tatoeba-test.eng-lad.tsv b/test/tatoeba-test.eng-lad.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f268a9f7b6702eb94615d50455eddf7a3137b14a
--- /dev/null
+++ b/test/tatoeba-test.eng-lad.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b71cfac1c7c849a443dc4a2dd21325eb38d5d6002f2a931cca413015f7fb0e0
+size 48380
diff --git a/test/tatoeba-test.eng-lad_Latn.tsv b/test/tatoeba-test.eng-lad_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..99f3a35ddfc54bdf1c21737e1983384b8b1ea9f7
--- /dev/null
+++ b/test/tatoeba-test.eng-lad_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d234354adf7502c344e39358e81d0bf6d427c4cef01482188863e06aaff0bfff
+size 40637
diff --git a/test/tatoeba-test.eng-lat.tsv b/test/tatoeba-test.eng-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f53d19ba738e1662fbe37820cacd8e8f32160c4c
--- /dev/null
+++ b/test/tatoeba-test.eng-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f2b7b8429b17127ebed72cae05503355493b57a359166bb0fcb97eee6090c8b
+size 968004
diff --git a/test/tatoeba-test.eng-lav.tsv b/test/tatoeba-test.eng-lav.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0e7291b3535c25d9c9266099f3f248d50ae7fa0c
--- /dev/null
+++ b/test/tatoeba-test.eng-lav.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:affa1731f394b49100070346a4c4e7d9bb5beaeb805ae5a359bfb08b5498b6a2
+size 114179
diff --git a/test/tatoeba-test.eng-lfn.tsv b/test/tatoeba-test.eng-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..25a75939eeec520a3984be62de10a86530e27967
--- /dev/null
+++ b/test/tatoeba-test.eng-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b9fa4d7dc605abb66370c73ec085ffdccfaab3106cf39f3d7fffe2abe836310
+size 263225
diff --git a/test/tatoeba-test.eng-lfn_Cyrl.tsv b/test/tatoeba-test.eng-lfn_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0c8be6df0343e33250f0b0532fcb1f6abb9b884f
--- /dev/null
+++ b/test/tatoeba-test.eng-lfn_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dde57725c9e0892fae70bba6c1a170426cddf1b116a93530218d946b5bff0b78
+size 79553
diff --git a/test/tatoeba-test.eng-lfn_Latn.tsv b/test/tatoeba-test.eng-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..25102012eeb482c740213f58a30b07376b84672c
--- /dev/null
+++ b/test/tatoeba-test.eng-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bceaaec14cdfe97e79d17c97a2c588f870a87cfeac51829cf549c5a77e06d57e
+size 183672
diff --git a/test/tatoeba-test.eng-lit.tsv b/test/tatoeba-test.eng-lit.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7bc4f5de96325c1dcb625e2ff8aba651b477fd5e
--- /dev/null
+++ b/test/tatoeba-test.eng-lit.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6431f6f9dcdf24cc532dd3e42a41fdb6273a53e5fa514a7150a70ecf9e811fd9
+size 180167
diff --git a/test/tatoeba-test.eng-ltz.tsv b/test/tatoeba-test.eng-ltz.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cf76c68f0a2d2739535ccf9081004bbc761f3e9f
--- /dev/null
+++ b/test/tatoeba-test.eng-ltz.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0b592496281877334fbaa298d834be2ea5fbc66d429a4196c6c7a14389cb018
+size 18648
diff --git a/test/tatoeba-test.eng-mal.tsv b/test/tatoeba-test.eng-mal.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2b5e0c5b6e8c757a1fb307d1e903dc8ddbc48857
--- /dev/null
+++ b/test/tatoeba-test.eng-mal.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2ee07901f46807fee391994c486821f30fc9bd251596b8b88d2df11b7de2a6a
+size 113410
diff --git a/test/tatoeba-test.eng-mar.tsv b/test/tatoeba-test.eng-mar.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bb5a7d7995f3f590684b8dcda6e5f034900f8f55
--- /dev/null
+++ b/test/tatoeba-test.eng-mar.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09ae4e17153e2e47f0bed0f626037cdecaef187dffc99ce6570f5e3a17b2a2db
+size 1139103
diff --git a/test/tatoeba-test.eng-mkd.tsv b/test/tatoeba-test.eng-mkd.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..421751a1e981a81a30c08c7eee3ae79be5983c84
--- /dev/null
+++ b/test/tatoeba-test.eng-mkd.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4978c5d1ddc2b3bcff62ed0b5e6f1b4d6856336113552716c1f7c0ecf0980f34
+size 863552
diff --git a/test/tatoeba-test.eng-mlt.tsv b/test/tatoeba-test.eng-mlt.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..09038562e99a9bd2886e1fc024dfc8bed62ae32a
--- /dev/null
+++ b/test/tatoeba-test.eng-mlt.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c399b8c7ab56c3f3212617f6d8270baddc660f1cc6cab891125171d75162fdc8
+size 11648
diff --git a/test/tatoeba-test.eng-mon.tsv b/test/tatoeba-test.eng-mon.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6cb30b5e780e9f1a99de87ec277ea2e43e94b569
--- /dev/null
+++ b/test/tatoeba-test.eng-mon.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0818c88bbbec1e656e65f31e993c011b1639fd5f3a53e41244eac428b48e7eee
+size 46279
diff --git a/test/tatoeba-test.eng-mri.tsv b/test/tatoeba-test.eng-mri.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e19c7268cf5e755ed5f9ae8ea250db01b83458ea
--- /dev/null
+++ b/test/tatoeba-test.eng-mri.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecdf024ba07463243872e4fdab7a990264ddd53806f2c02c7e9aec0f9e45bdb8
+size 30911
diff --git a/test/tatoeba-test.eng-msa.tsv b/test/tatoeba-test.eng-msa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5545b6b802152953dc08ce5f2e9f898eb8a7a241
--- /dev/null
+++ b/test/tatoeba-test.eng-msa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8d6d39a62bf8968d0d865fe3c99abc181cb55823b71dfd89cc15e1a7d042410
+size 392371
diff --git a/test/tatoeba-test.eng-mya.tsv b/test/tatoeba-test.eng-mya.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..19fbf50b0ec68ff17e2a876f4197eb355527d9d0
--- /dev/null
+++ b/test/tatoeba-test.eng-mya.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbf22f3cf0c81a1334f730a365d11faae86313b7380c63fbfe878a80867d2910
+size 34381
diff --git a/test/tatoeba-test.eng-nds.tsv b/test/tatoeba-test.eng-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3b9518652171c43518ba984927a0acb257d453cf
--- /dev/null
+++ b/test/tatoeba-test.eng-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b50bb599ee48fb77ed9944d74f82c130f12bd54b6cb16ff38096f279e08e530f
+size 178982
diff --git a/test/tatoeba-test.eng-nld.tsv b/test/tatoeba-test.eng-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e415755574926fc117e21b63665062220c4c2e4d
--- /dev/null
+++ b/test/tatoeba-test.eng-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce0f59ef9049ddb2b7e4a728e74779d81e408363aef89c13bb7159142857cc5d
+size 936676
diff --git a/test/tatoeba-test.eng-nno.tsv b/test/tatoeba-test.eng-nno.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..163d90a706fbf6c2c67b881508b8a84a70cc91ef
--- /dev/null
+++ b/test/tatoeba-test.eng-nno.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f3702e1ce3dff2008ba4343b15fac7233e2630067601c8bf63aea1f265cefa2
+size 34843
diff --git a/test/tatoeba-test.eng-nob.tsv b/test/tatoeba-test.eng-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8bdad8ee55766d5ea9c9e9b4a54b4fc78c29031f
--- /dev/null
+++ b/test/tatoeba-test.eng-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a886e9887803623bae6da56ce543d67de58ac6bd01dd9186b8f044382e818e42
+size 370388
diff --git a/test/tatoeba-test.eng-nor.tsv b/test/tatoeba-test.eng-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5716dedd328ceedbc46d7cf26edf464b3190e281
--- /dev/null
+++ b/test/tatoeba-test.eng-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc5d3f9979a1fde6ab991cfb22f7c993cd01528659d228ec04a53bcf01dc2f2f
+size 405307
diff --git a/test/tatoeba-test.eng-nov.tsv b/test/tatoeba-test.eng-nov.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d3102830d4cb29d0f2a638c0764c35c347ab5a70
--- /dev/null
+++ b/test/tatoeba-test.eng-nov.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2be7738741a3c7904da8bd3ef524b4a4c57a162be6f1dac49b3cf68afb9640a
+size 14871
diff --git a/test/tatoeba-test.eng-nst.tsv b/test/tatoeba-test.eng-nst.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ccff866a84e05d9f28344dfb4fa2c4ddcb567fb5
--- /dev/null
+++ b/test/tatoeba-test.eng-nst.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d95ee1ece3e0eee0d8cfd10608808c73b09f24c0cb41b7166e1931dcdbdbf3a
+size 61426
diff --git a/test/tatoeba-test.eng-oci.tsv b/test/tatoeba-test.eng-oci.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d71ca62b871464f985ef33fa6a742e1c1b0fff0d
--- /dev/null
+++ b/test/tatoeba-test.eng-oci.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45c0533fae9a66452786619d83e02e949c7c997f0d0662055341ae0695150c13
+size 54372
diff --git a/test/tatoeba-test.eng-orv.tsv b/test/tatoeba-test.eng-orv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9f221553595683fca0b30666b563a4ef63c3ef46
--- /dev/null
+++ b/test/tatoeba-test.eng-orv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57b0d9b20ad1a3cf535b1522576433f30e8fc46e001fdda78990df3969ffaf4e
+size 26658
diff --git a/test/tatoeba-test.eng-ota.tsv b/test/tatoeba-test.eng-ota.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..31e625921f66f563a7aa0daaf2447f310e3c68cf
--- /dev/null
+++ b/test/tatoeba-test.eng-ota.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:931a22eb31aad94cbe729035bd9894134688ecbae34b7c4ec03a9ff78dfdc8c5
+size 58713
diff --git a/test/tatoeba-test.eng-ota_Arab.tsv b/test/tatoeba-test.eng-ota_Arab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0bc1f9d69b2a2c4a8fa5782e4b76ed87d6b8125f
--- /dev/null
+++ b/test/tatoeba-test.eng-ota_Arab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:135519103253ccbb387e0f74a16896e2dc2544f8d4f9869eae6488e5cf6afc4f
+size 33731
diff --git a/test/tatoeba-test.eng-ota_Latn.tsv b/test/tatoeba-test.eng-ota_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ec5e92cb03a3793598087533b55a40a005e7eb18
--- /dev/null
+++ b/test/tatoeba-test.eng-ota_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39a8a91def911c407d788c67057d8e61bc0ef73c848d9baa54b0f7defebe1dec
+size 24982
diff --git a/test/tatoeba-test.eng-pam.tsv b/test/tatoeba-test.eng-pam.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d24984fbbc81225c6030a605c9fe72dab37c96b5
--- /dev/null
+++ b/test/tatoeba-test.eng-pam.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b6b59c67e5a4b4a73fc901ca9abcf4d9d0716a1aef4342110b30159ccc229b8
+size 75041
diff --git a/test/tatoeba-test.eng-pes.tsv b/test/tatoeba-test.eng-pes.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e4da335c18fbd7b359944a5b86b7e229a9183275
--- /dev/null
+++ b/test/tatoeba-test.eng-pes.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a67871820da1c1b6cc3579d9c3fdb733ddc5aa79a82e7c7e8e70faed98c7c847
+size 409230
diff --git a/test/tatoeba-test.eng-pms.tsv b/test/tatoeba-test.eng-pms.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..40d5d007671711986a26ecf023e06811f4a86666
--- /dev/null
+++ b/test/tatoeba-test.eng-pms.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5949d2e7f0fba9ce9e6e4c4aa7bb2644a3de0e734768e28c71c53277e4013d9
+size 21102
diff --git a/test/tatoeba-test.eng-pol.tsv b/test/tatoeba-test.eng-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6277828df844bf496498193b665f860048eb7aa1
--- /dev/null
+++ b/test/tatoeba-test.eng-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4485fb0e52f5275b93b3e07811a1381c455a8b0245b86507e5f1b2f1b72e4050
+size 780758
diff --git a/test/tatoeba-test.eng-por.tsv b/test/tatoeba-test.eng-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2f2fdc19d3398c85147610b0703f99b94f339b73
--- /dev/null
+++ b/test/tatoeba-test.eng-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3cd74ff832a693613d98676a22237cf1969724978229d1c815ad7c989d2cf06
+size 1085168
diff --git a/test/tatoeba-test.eng-prg.tsv b/test/tatoeba-test.eng-prg.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8a0a3a1f8f00002f93bcc9b54f09b4ff3a301814
--- /dev/null
+++ b/test/tatoeba-test.eng-prg.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09f271d7658da1156a3c0f14cdc94c284bd2c72e682e0a5a58b30bc57cb7f12e
+size 19065
diff --git a/test/tatoeba-test.eng-que.tsv b/test/tatoeba-test.eng-que.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ac44d7da931dc4aadaa1eaa862911d1c28e37e45
--- /dev/null
+++ b/test/tatoeba-test.eng-que.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c3433b346d3a08ba1b8b35e79e160b5effa017089f54bc86f991fb5fd480467
+size 15806
diff --git a/test/tatoeba-test.eng-rom.tsv b/test/tatoeba-test.eng-rom.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1f126a637f4b4b02519e4204cc37d1cbfa0d5c4b
--- /dev/null
+++ b/test/tatoeba-test.eng-rom.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:401cec67da830bc24d0d49b536df5f4d95a5f80089a06266b0ad6732409016f0
+size 48250
diff --git a/test/tatoeba-test.eng-ron.tsv b/test/tatoeba-test.eng-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..43460cae19737d2039310314ea112fa21d1baafb
--- /dev/null
+++ b/test/tatoeba-test.eng-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a62b08f978616e399de570594b16ae62e1323a9df7f52315deab5a42c2037c10
+size 429293
diff --git a/test/tatoeba-test.eng-run.tsv b/test/tatoeba-test.eng-run.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d7b81b3a70d604ed76e0d344f5167ba932fa45eb
--- /dev/null
+++ b/test/tatoeba-test.eng-run.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a218dd41458340da3d64639a12b3b85fd1861147ea8eefe7730fdd17e27df3b2
+size 90358
diff --git a/test/tatoeba-test.eng-rus.tsv b/test/tatoeba-test.eng-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0ff2ae4b8a80b198e89b1db30306d84af818f54a
--- /dev/null
+++ b/test/tatoeba-test.eng-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6a8d4550fccd75ff610e8b15d9f499a78230b610757799d1b789569c3521c89
+size 1947659
diff --git a/test/tatoeba-test.eng-slv.tsv b/test/tatoeba-test.eng-slv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3dbe92da8833dfbc68c2f8a91c1251fbc7ee143b
--- /dev/null
+++ b/test/tatoeba-test.eng-slv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eeaece101a2ee2da97e1585f09711af0cbe4bca5af0665f8864eeab7e055ce00
+size 166748
diff --git a/test/tatoeba-test.eng-spa.tsv b/test/tatoeba-test.eng-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..78db8879cfddba4de30c019b3540f9b575c75925
--- /dev/null
+++ b/test/tatoeba-test.eng-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74c9e2d59426f6b1bfe92c6e6a9f0cb1847e95e06e114ca9f38d656046253752
+size 1420922
diff --git a/test/tatoeba-test.eng-sqi.tsv b/test/tatoeba-test.eng-sqi.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..18d48cf97b787395ac55e7a60329c6248375f993
--- /dev/null
+++ b/test/tatoeba-test.eng-sqi.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11442f9392c3ec524e8d2ddac7e7a73698dd3b180944931dc4af0c4b5c6e72c4
+size 85651
diff --git a/test/tatoeba-test.eng-srp_Cyrl.tsv b/test/tatoeba-test.eng-srp_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..29f7a9c3c1398427eb40e4d238e5b88faeeea170
--- /dev/null
+++ b/test/tatoeba-test.eng-srp_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5ff36ae480165205c828d6b8a7c49f172c02cf733c373fd9ea6f0df64370bc7
+size 136541
diff --git a/test/tatoeba-test.eng-srp_Latn.tsv b/test/tatoeba-test.eng-srp_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a3ad14afdfba2b30d891e08495f3d049721171f6
--- /dev/null
+++ b/test/tatoeba-test.eng-srp_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2b751480dfa3a4523651393d19fb27d159ccbee61079319aacabacb03bd4a3b
+size 497008
diff --git a/test/tatoeba-test.eng-swa.tsv b/test/tatoeba-test.eng-swa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..059e218dde2fbcedf63b3edfec7ace1e442320eb
--- /dev/null
+++ b/test/tatoeba-test.eng-swa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c1e5737bba9191b09e34b6205d5009a2d2f7454c76f0386e7ef6b500a00c301
+size 22845
diff --git a/test/tatoeba-test.eng-swe.tsv b/test/tatoeba-test.eng-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..58b24d5c30347ce8f5fca5a97dc1f5094c8952e5
--- /dev/null
+++ b/test/tatoeba-test.eng-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3344951f7eb060d46a66731e52de6a0c41970c421d80e2cbedc706e47607e3b
+size 697087
diff --git a/test/tatoeba-test.eng-tam.tsv b/test/tatoeba-test.eng-tam.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..38e931f5cc8e4e9487e30c894df4dda977fed502
--- /dev/null
+++ b/test/tatoeba-test.eng-tam.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ab7ecb33be3caa6e602cc9481e6217eb56b7e6f6179d18ce9d4643a3d4e3096
+size 41917
diff --git a/test/tatoeba-test.eng-tat.tsv b/test/tatoeba-test.eng-tat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6ac996181804a4a7615dfe43029bf3d5509dc263
--- /dev/null
+++ b/test/tatoeba-test.eng-tat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1735dcf69bfd93930af07c42d3fa320d8f737c99d0702a7d189ab3943330051
+size 143755
diff --git a/test/tatoeba-test.eng-tel.tsv b/test/tatoeba-test.eng-tel.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..18cdcc9208a08458aa2c79301dcd7a220a54aa28
--- /dev/null
+++ b/test/tatoeba-test.eng-tel.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df531d7f01e25506579f02a5316a85f8fd7d18baeaaf7b36cbdef648855f0e16
+size 28967
diff --git a/test/tatoeba-test.eng-tgl.tsv b/test/tatoeba-test.eng-tgl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2d1a6166e478d63ef2751dfacbfc2e0127e01131
--- /dev/null
+++ b/test/tatoeba-test.eng-tgl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e951f41e6b7cf59bf656c37c44a467df6aa045bc1915c3597b57662f5e3d87bf
+size 202000
diff --git a/test/tatoeba-test.eng-tha.tsv b/test/tatoeba-test.eng-tha.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4b112ec9a18e4486b82289a13303513da91969fd
--- /dev/null
+++ b/test/tatoeba-test.eng-tha.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd9b1b709bde693507d7858de831746e31b706cd85aca09012e16a7706d0fda7
+size 120449
diff --git a/test/tatoeba-test.eng-tlh.tsv b/test/tatoeba-test.eng-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c2db20c1f10239ffa455e330bed8460a7e78de4f
--- /dev/null
+++ b/test/tatoeba-test.eng-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b07189d019829a7c58dd8d6d467dbde337f48b784b5dbcd3f1c31bde9fb6ea37
+size 321225
diff --git a/test/tatoeba-test.eng-toki.tsv b/test/tatoeba-test.eng-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2d61fc942f6ba9af2dce4004bf7373d3516a4608
--- /dev/null
+++ b/test/tatoeba-test.eng-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4cf98dd68ab2911495778d57ed15002b879a5de82980b2595cea9affe269f12a
+size 401418
diff --git a/test/tatoeba-test.eng-tuk.tsv b/test/tatoeba-test.eng-tuk.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..15be8991f830c3e43b0adfa85af8a5998e1e57f4
--- /dev/null
+++ b/test/tatoeba-test.eng-tuk.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29b54bbdab53ee9f00b7c21a5e90f3ea4f9cc31dae1f899ef8eeba1eb9ee60d2
+size 213395
diff --git a/test/tatoeba-test.eng-tuk_Latn.tsv b/test/tatoeba-test.eng-tuk_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a26d35931fa8e4282df9d917b8a3b607af3cc1d1
--- /dev/null
+++ b/test/tatoeba-test.eng-tuk_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:017b8bc5913afc120b13f149b7b9bcf606be528dbc9eae909aac55af40a5418d
+size 213379
diff --git a/test/tatoeba-test.eng-tur.tsv b/test/tatoeba-test.eng-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0f72b3b4ce7bd3959113b7e6cacd781f9c87c9bd
--- /dev/null
+++ b/test/tatoeba-test.eng-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d305fd0e5b65710a5c43864ba43cee6d25ff010ade6402efb6aae254e96441c1
+size 1133690
diff --git a/test/tatoeba-test.eng-tzl.tsv b/test/tatoeba-test.eng-tzl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fc735d7ac28f8f100cbeafab24d6b5c2aa0cf5e8
--- /dev/null
+++ b/test/tatoeba-test.eng-tzl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:323969b0814101fc465f081fd35c2bd75c86f691e8ebee58448f3822399694af
+size 10369
diff --git a/test/tatoeba-test.eng-tzl_Latn.tsv b/test/tatoeba-test.eng-tzl_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2d8e50b8c18ab05eac0ce6d749d367b62dc38bd5
--- /dev/null
+++ b/test/tatoeba-test.eng-tzl_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d956a6e85888e75d013edbd494748e6e56792f2eb0cbcb2201c9ce2cf1f490d0
+size 10349
diff --git a/test/tatoeba-test.eng-uig.tsv b/test/tatoeba-test.eng-uig.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8270abb54549334dabfa212754d8f9d4601d7a7e
--- /dev/null
+++ b/test/tatoeba-test.eng-uig.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c37fdef83783568547f882a8f0661dcc215fad58c38c5d0cec9dfa11ae318de9
+size 314767
diff --git a/test/tatoeba-test.eng-uig_Arab.tsv b/test/tatoeba-test.eng-uig_Arab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cbb9f18350ee1349d762a146ec21ee49ea65e8c0
--- /dev/null
+++ b/test/tatoeba-test.eng-uig_Arab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8754d2c36a00786df2b186d07f8fd9d12b9213f7f8415c7604ee9cdc0aef87a
+size 314408
diff --git a/test/tatoeba-test.eng-ukr.tsv b/test/tatoeba-test.eng-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b427b93b1f1dee3ffdd50ab5d66b3f2bf0557a5e
--- /dev/null
+++ b/test/tatoeba-test.eng-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f632f1830b9153980074ea507d69e8000eeae05535df78c6783cce8199ee474
+size 1165759
diff --git a/test/tatoeba-test.eng-urd.tsv b/test/tatoeba-test.eng-urd.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7797c7aec12e55798fb47993fb8022e30eaa1090
--- /dev/null
+++ b/test/tatoeba-test.eng-urd.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c722147f0dfd2b137ae1284832717eb6f1213a5f579feb07920964956db67bab
+size 160779
diff --git a/test/tatoeba-test.eng-uzb.tsv b/test/tatoeba-test.eng-uzb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5ef1438480dc957a45ca80067332e42729f50ee6
--- /dev/null
+++ b/test/tatoeba-test.eng-uzb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4559e8c2f02d0eec2216743b7303880f9742c8eb26bf6d5e39ebd51ffec543c1
+size 31244
diff --git a/test/tatoeba-test.eng-uzb_Latn.tsv b/test/tatoeba-test.eng-uzb_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3d3dbf30d4064025a5ccca32c5cef50094512dd4
--- /dev/null
+++ b/test/tatoeba-test.eng-uzb_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb15f41c871b398df0ccd668249b04227b05870ca4fda74df4652d8a8d4821e8
+size 17949
diff --git a/test/tatoeba-test.eng-vie.tsv b/test/tatoeba-test.eng-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2522bd315571f076d327cf20dcd680e64e872cc9
--- /dev/null
+++ b/test/tatoeba-test.eng-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c644c5d785cfccdc77cea9a0bbded7b6e20dcf17fd90295b0d646e2d7b867ec1
+size 234244
diff --git a/test/tatoeba-test.eng-vol.tsv b/test/tatoeba-test.eng-vol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b968565c062f2b7d8274d3ba111ce75c58b0a217
--- /dev/null
+++ b/test/tatoeba-test.eng-vol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fe472971072873491df81ff8c1e2a3624c098853493aec2bc0d6fd372c0f47e
+size 100559
diff --git a/test/tatoeba-test.eng-war.tsv b/test/tatoeba-test.eng-war.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..eb8b16c7984643620e5bd635955587abad2b44e2
--- /dev/null
+++ b/test/tatoeba-test.eng-war.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f89ae0f1dd8adea5fc483d50f560c1b3b3968d6f596c36fb6f982581c7d4c1bc
+size 117917
diff --git a/test/tatoeba-test.eng-xal.tsv b/test/tatoeba-test.eng-xal.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2f6b6792506d24436aab6b82e1df568b00a586c9
--- /dev/null
+++ b/test/tatoeba-test.eng-xal.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18d5796bb080fe847f9a1553070509ed894b4e063a17d19b291cf284c48bf64d
+size 25020
diff --git a/test/tatoeba-test.eng-yid.tsv b/test/tatoeba-test.eng-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b1be6abeab0a424b4359a7392bba2bf2e6d84f95
--- /dev/null
+++ b/test/tatoeba-test.eng-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87229275d09978896bcf140856362dfd064bcdd015ecb0ae7579ae3b533e42a5
+size 232158
diff --git a/test/tatoeba-test.eng-yue_Hans.tsv b/test/tatoeba-test.eng-yue_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..903ac5f9ecdfb5b1d75004fb6995f36ba5e63879
--- /dev/null
+++ b/test/tatoeba-test.eng-yue_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d19d335a9cbf4eb08f7db5c86fe67edd7ab9f44937fea6e1d8a54e5545475826
+size 63715
diff --git a/test/tatoeba-test.eng-yue_Hant.tsv b/test/tatoeba-test.eng-yue_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bdfa65ad6e9668e8e551d71dd99db7cefdbaac47
--- /dev/null
+++ b/test/tatoeba-test.eng-yue_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:754fdb327a2524926bafba702cea1cb841a9fe6f75eede219b7afe786fee7f96
+size 35124
diff --git a/test/tatoeba-test.eng-zho.tsv b/test/tatoeba-test.eng-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bf5c94f9cdd444a7cadbb8349881de30d60269cd
--- /dev/null
+++ b/test/tatoeba-test.eng-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06f9a8a04b92e0d7b2e8441f05de9952c6b31fba2dcae834d7cba03190abd284
+size 871802
diff --git a/test/tatoeba-test.eng-zsm_Latn.tsv b/test/tatoeba-test.eng-zsm_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7ed4aa4032d145125c06f176a59b0735b4975ad7
--- /dev/null
+++ b/test/tatoeba-test.eng-zsm_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6ed126a7d43bac7ff29c8af50dbfbc6024625b4c7448aa1336e58dc9426f590
+size 49436
diff --git a/test/tatoeba-test.eng-zza.tsv b/test/tatoeba-test.eng-zza.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b690cf3c4eaf563983d6e3130274a050f1709f39
--- /dev/null
+++ b/test/tatoeba-test.eng-zza.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68b8be79349e00f84409d26fe771ba4e4d4dbf367310c924fc927659484b32d8
+size 29734
diff --git a/test/tatoeba-test.epo-cmn_Hans.tsv b/test/tatoeba-test.epo-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8c85782db11686d9ec1582b85e2ee392d489f933
--- /dev/null
+++ b/test/tatoeba-test.epo-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f573d61b028777ddf6d91943afb8f6259ab607dd2bf6ee97faa44fdd43b211dc
+size 111383
diff --git a/test/tatoeba-test.epo-cmn_Hant.tsv b/test/tatoeba-test.epo-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..312030948204e816edc7a1a9a09337f67452a6e8
--- /dev/null
+++ b/test/tatoeba-test.epo-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b2f05ef49f6a636422f3a208c2084afa4093968e7b371e8698cd5353df7485a
+size 60824
diff --git a/test/tatoeba-test.epo-epo.tsv b/test/tatoeba-test.epo-epo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7447e9aae38b514f1a44c4f9df38a59c6424131b
--- /dev/null
+++ b/test/tatoeba-test.epo-epo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb247407b10eeccb4eae3bde995489281427f4d4dfc1256c466efbca45c22c70
+size 797530
diff --git a/test/tatoeba-test.epo-fas.tsv b/test/tatoeba-test.epo-fas.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6eb8998749efaee1b01b761d5b45c7d9727a01b1
--- /dev/null
+++ b/test/tatoeba-test.epo-fas.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a505c3cbdd7985cbb9238fdc08eb6da4a8e5da6c2b6ee8ce8fa18b9cb5bd19f
+size 268962
diff --git a/test/tatoeba-test.epo-fin.tsv b/test/tatoeba-test.epo-fin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9b1f2e277767cd0e65e40effa1cb4461be216930
--- /dev/null
+++ b/test/tatoeba-test.epo-fin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e4a32d8a9e5b75d6f3fa46aa54331f4c564faf73628c62b879fea6ce244d0bb
+size 206658
diff --git a/test/tatoeba-test.epo-fra.tsv b/test/tatoeba-test.epo-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..995041f2d689d338b4e0f027f0ce0be82f338b6e
--- /dev/null
+++ b/test/tatoeba-test.epo-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9339a59239175d0317a00f0ba3a39f67667a73aa1d730186ac6c87d9ca47206
+size 1084431
diff --git a/test/tatoeba-test.epo-glg.tsv b/test/tatoeba-test.epo-glg.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..aecbb1daae077ad8e2c44b83db1baf9b2630e56b
--- /dev/null
+++ b/test/tatoeba-test.epo-glg.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55314e36224b641e3ceadae06f75cfaee03d1ae0ea6ab5f98e7bc1f11ebd9eef
+size 24476
diff --git a/test/tatoeba-test.epo-hbs.tsv b/test/tatoeba-test.epo-hbs.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0ae978fbb2f6379e83146766ac8755a6fedd33c5
--- /dev/null
+++ b/test/tatoeba-test.epo-hbs.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe9c55cb94bfd11b77db2a2176d7e41c8e9b7f49e77c5c285668d6ca90702945
+size 258348
diff --git a/test/tatoeba-test.epo-heb.tsv b/test/tatoeba-test.epo-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ca1ebb4cf0cae7bb1fd699954985b973532c061f
--- /dev/null
+++ b/test/tatoeba-test.epo-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:accee6385c6dbdd0c0376c5b3b3d004a53ee83a1d26a5ecdc16f90d5db073c1b
+size 988067
diff --git a/test/tatoeba-test.epo-hrv.tsv b/test/tatoeba-test.epo-hrv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..945a4c62b762174df49cea484d681054dd443589
--- /dev/null
+++ b/test/tatoeba-test.epo-hrv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40d28b2f3ffd4f984236897f6435d6754380f3dd77cc0bbf0f773ff65665f2b1
+size 33262
diff --git a/test/tatoeba-test.epo-hun.tsv b/test/tatoeba-test.epo-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0839762703c0d7633d3756990269db0eb3b742f6
--- /dev/null
+++ b/test/tatoeba-test.epo-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8415794602bc2969316f103009bb6a65b56dfbcfb7f62a9e0775e807a3341599
+size 843261
diff --git a/test/tatoeba-test.epo-ido.tsv b/test/tatoeba-test.epo-ido.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..06f6362c346b40fba8ec05990ab1fc69e4e219e3
--- /dev/null
+++ b/test/tatoeba-test.epo-ido.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:715ff4e93ea6ef5ad436cfa9c1a8dcfe47fb8409b145fb1fd8472595d4e9dc36
+size 87309
diff --git a/test/tatoeba-test.epo-ile.tsv b/test/tatoeba-test.epo-ile.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..16cfb57015b9a3a7a44e06c6bd2a515a1abe8da6
--- /dev/null
+++ b/test/tatoeba-test.epo-ile.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94bc2e7928f8d4bacb14a45410a47687d658c297a1bc253ac539afec22b68ee6
+size 19439
diff --git a/test/tatoeba-test.epo-ile_Latn.tsv b/test/tatoeba-test.epo-ile_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7d8a59197f7cedd4accbd10042042fae86bf2338
--- /dev/null
+++ b/test/tatoeba-test.epo-ile_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1727f224801a6f3ddbe232edc537686f6bcde8f941efe0648f933f24976171ca
+size 19423
diff --git a/test/tatoeba-test.epo-ina.tsv b/test/tatoeba-test.epo-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cc7e23de17533259c91bc5fb6a3b41fe6a3d7455
--- /dev/null
+++ b/test/tatoeba-test.epo-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f7406de093343a6b2a9b76d02c60570367bb6f1419ab9ad87d9bd0609ad8e38
+size 250841
diff --git a/test/tatoeba-test.epo-isl.tsv b/test/tatoeba-test.epo-isl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cc6835b014f116869d3a3bf1081e45d51b9421f9
--- /dev/null
+++ b/test/tatoeba-test.epo-isl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d19a3049cc719abcc1be14103d4894a32bf59c536f6e61fcc4b855e8ac336219
+size 17137
diff --git a/test/tatoeba-test.epo-ita.tsv b/test/tatoeba-test.epo-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..332325514afed1466865ac11b6ef9a1ed689fe27
--- /dev/null
+++ b/test/tatoeba-test.epo-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fecf2b8359cb33fd475ac34ce796946a37c87981e424ed45b64d65d676d534b7
+size 765268
diff --git a/test/tatoeba-test.epo-jbo.tsv b/test/tatoeba-test.epo-jbo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..35ccf00944f583add71416c21ef303f90aea12b8
--- /dev/null
+++ b/test/tatoeba-test.epo-jbo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57f54dadf6d163e8a68dc09d3eb26228cb4bcffea8bc0ee2ee7b1300b68dac88
+size 83091
diff --git a/test/tatoeba-test.epo-jpn.tsv b/test/tatoeba-test.epo-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..066d88955649a8838658246ed878a6676b7db866
--- /dev/null
+++ b/test/tatoeba-test.epo-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72019ab045b517b5e5199f7064748fb8561a676de39eebf1e94009f46ce9dc90
+size 334962
diff --git a/test/tatoeba-test.epo-jpn_Hani.tsv b/test/tatoeba-test.epo-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d20e24dec165673520652534a6bce0c7af4d5fca
--- /dev/null
+++ b/test/tatoeba-test.epo-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31913e69f96a92e8f91a539602b36afdf40d8ab0325f7e1c79e09b33f03ab9a7
+size 31274
diff --git a/test/tatoeba-test.epo-jpn_Hira.tsv b/test/tatoeba-test.epo-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9bf5c20d3acf4a62c6a3e21d0b2f89f5573fda95
--- /dev/null
+++ b/test/tatoeba-test.epo-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3758c5ae2309430b79e885fc943d7992f5b5280c1729ba36a0f0ec41655df685
+size 289591
diff --git a/test/tatoeba-test.epo-lad.tsv b/test/tatoeba-test.epo-lad.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d9b371846a16ae22a034961e6156e37b28694b4c
--- /dev/null
+++ b/test/tatoeba-test.epo-lad.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0ee55268ad5c601af546ed8d3ac8b22b60bfa5d0fe958c169a6f048730c7af7
+size 25590
diff --git a/test/tatoeba-test.epo-lad_Latn.tsv b/test/tatoeba-test.epo-lad_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a46cb1832700640a9e3104a3293f7506386093e4
--- /dev/null
+++ b/test/tatoeba-test.epo-lad_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58d4250c6b6d64656ebed964f6e27daebbc670ce9e6832d09c22c253483f3891
+size 17121
diff --git a/test/tatoeba-test.epo-lat.tsv b/test/tatoeba-test.epo-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..554c1fcd3b5bbe3da72da1f88bb908c3c57d98aa
--- /dev/null
+++ b/test/tatoeba-test.epo-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcd0245cdfda9cb79370d5ad95c6bb818ec87a50a1124f08b29fd50973acded8
+size 395552
diff --git a/test/tatoeba-test.epo-lfn.tsv b/test/tatoeba-test.epo-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..99f3f7d6a58c7967964cef6605266ee1b4314f44
--- /dev/null
+++ b/test/tatoeba-test.epo-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7625810cbfc7ed7bb7abff8616654e6a41cd5c3fdde0fea8f2ac19a51098eac8
+size 73629
diff --git a/test/tatoeba-test.epo-lfn_Latn.tsv b/test/tatoeba-test.epo-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5f583f29d99833676c6bc9950f2af9cf5fdf914c
--- /dev/null
+++ b/test/tatoeba-test.epo-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:663b62d64b417e27bbfb65b206e3aa926ed1f73f49691a3602ec176053c8ca62
+size 60368
diff --git a/test/tatoeba-test.epo-lit.tsv b/test/tatoeba-test.epo-lit.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b8248ae11ce8db3f4a7ce262a974e7961f50f526
--- /dev/null
+++ b/test/tatoeba-test.epo-lit.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71adf19565623db1c5545699efced4c06d6a5e33003c8ebaa86d20640e8935ff
+size 879872
diff --git a/test/tatoeba-test.epo-nds.tsv b/test/tatoeba-test.epo-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..29139b2f5b8b0fe2867ba441b09bcca264b2c188
--- /dev/null
+++ b/test/tatoeba-test.epo-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bf105898b5a42f5c1efd056bebd47c6e8a73002dacde14705caaf8b71e8b796
+size 183770
diff --git a/test/tatoeba-test.epo-nld.tsv b/test/tatoeba-test.epo-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..465a495a6490a4cee3e288004b7669b491e53ccb
--- /dev/null
+++ b/test/tatoeba-test.epo-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14eed8014f73dd4e513f433076b2ba657832da90446469fb2e5548eb98ccf248
+size 1024905
diff --git a/test/tatoeba-test.epo-nob.tsv b/test/tatoeba-test.epo-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..08daaad6da99db0b266a2eb5691f5ede7d7a3de2
--- /dev/null
+++ b/test/tatoeba-test.epo-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf20565ef94c9d6a39656b97982958f07864f844734f690b8add4f91d6140fe6
+size 308406
diff --git a/test/tatoeba-test.epo-nor.tsv b/test/tatoeba-test.epo-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..44b59becac3b189bafe372e879dfed2934f9dbc3
--- /dev/null
+++ b/test/tatoeba-test.epo-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ff42de612a42aa12646b8754ff88d5cc219866f53cd9072c3199496cc71c191
+size 313130
diff --git a/test/tatoeba-test.epo-oci.tsv b/test/tatoeba-test.epo-oci.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a17c92cc9f26be1b42799f89ae9d45605d496009
--- /dev/null
+++ b/test/tatoeba-test.epo-oci.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28758a3d988c5f93bb45f28af15a638ed972e014d0d7a45aa851a75ea734d396
+size 54830
diff --git a/test/tatoeba-test.epo-pol.tsv b/test/tatoeba-test.epo-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..eeedd2fcb2bb6e2260fe2ae58df7a83ab03fc61e
--- /dev/null
+++ b/test/tatoeba-test.epo-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c05ee4403354c9e46341b0a58be530855709376ae218d36741c389230b0e9c4
+size 190799
diff --git a/test/tatoeba-test.epo-por.tsv b/test/tatoeba-test.epo-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c78e2b8f3eeca5a046ad6f0a6a2f59cebd72252c
--- /dev/null
+++ b/test/tatoeba-test.epo-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8eee557cacfb158708d697d9441d365ce9b2bba3a72f27672ccd02b41023a0fc
+size 1038541
diff --git a/test/tatoeba-test.epo-ron.tsv b/test/tatoeba-test.epo-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..af27b986dc3d1aeac682cfb1760475c0f647f276
--- /dev/null
+++ b/test/tatoeba-test.epo-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb222277dea31e765ed7c906d3c0abb1244c8f9a8caba719d1bf9785752931df
+size 279807
diff --git a/test/tatoeba-test.epo-rus.tsv b/test/tatoeba-test.epo-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a51106610f71fcd2ca99d917712394d54f4d9f81
--- /dev/null
+++ b/test/tatoeba-test.epo-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c825457aa3a58b23f31275846fc0f1ab9af5f50130ab171e805403a5eee9d739
+size 1248575
diff --git a/test/tatoeba-test.epo-slv.tsv b/test/tatoeba-test.epo-slv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..336db0bfd26952f7a1aa499b4b8ddc4bbfeb9720
--- /dev/null
+++ b/test/tatoeba-test.epo-slv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:506b7dc3868c63ee66dfac0151b105ba8f532877c11082532e2a22f0dae02fee
+size 23153
diff --git a/test/tatoeba-test.epo-spa.tsv b/test/tatoeba-test.epo-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6941c3c6c4cd46b06b70f074861f01b178a3d8db
--- /dev/null
+++ b/test/tatoeba-test.epo-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fe4e7bc0f67469bf31cbf924b4be0fea6f7fdae9d12beb52886d0a0a759453e
+size 894253
diff --git a/test/tatoeba-test.epo-srp_Cyrl.tsv b/test/tatoeba-test.epo-srp_Cyrl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f0976d664785ba07aebec0b0782eb8c0c5cc5389
--- /dev/null
+++ b/test/tatoeba-test.epo-srp_Cyrl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc1c4aee122017c2fa21b451b7f569ec84b62150319ad83e31e14faab4f24304
+size 203501
diff --git a/test/tatoeba-test.epo-srp_Latn.tsv b/test/tatoeba-test.epo-srp_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ac9810c34ca16512bbc63832f0b6e41b795c6f17
--- /dev/null
+++ b/test/tatoeba-test.epo-srp_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e23e9f2dcca9ea3e1c44bc88275d14d2deb582e4604860a8cbe1c4b286276c5
+size 20921
diff --git a/test/tatoeba-test.epo-swe.tsv b/test/tatoeba-test.epo-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8f588a366ceb80a3c9712882ba0c8cd21028d879
--- /dev/null
+++ b/test/tatoeba-test.epo-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d0795b429ff9e396c1939453f75f9e471728c6e709dece77f6d39514723afdf
+size 119716
diff --git a/test/tatoeba-test.epo-tgl.tsv b/test/tatoeba-test.epo-tgl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..586b2b64c804b467c1b9d0cf7a3b39a03024a750
--- /dev/null
+++ b/test/tatoeba-test.epo-tgl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a117cc678ce882707a9285cddcdc3fa996e902b14230f559eec39408d471dbc
+size 101982
diff --git a/test/tatoeba-test.epo-tlh.tsv b/test/tatoeba-test.epo-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..03af2d2a7e8effd0e1a8e8cb4a15db52b8df294d
--- /dev/null
+++ b/test/tatoeba-test.epo-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d40f9871355971122b3a1909c2947736303b57fa09106c0dec5d6e7a9f17c11f
+size 127748
diff --git a/test/tatoeba-test.epo-toki.tsv b/test/tatoeba-test.epo-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7f9f7531d93699be7b91e232f98252800ee01a0d
--- /dev/null
+++ b/test/tatoeba-test.epo-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbfe57d6d3886b0f2bc2ed9add078778e2576b0fdc113345d7d36fc736b841e7
+size 234998
diff --git a/test/tatoeba-test.epo-tur.tsv b/test/tatoeba-test.epo-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f64e8d7b4b41b4c64e29f153e508a4abe7d4273d
--- /dev/null
+++ b/test/tatoeba-test.epo-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a76afe1a6231bd100b0605ee70cc38eb7f047690473bac38de3e53b7de5b3ffa
+size 359673
diff --git a/test/tatoeba-test.epo-ukr.tsv b/test/tatoeba-test.epo-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..efbb823ac42baca9fce16486884a60906f26dee1
--- /dev/null
+++ b/test/tatoeba-test.epo-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ddb40dfa0cb05486e4464d67eae37ae03d4124877115c2ae1c37e6251c098098
+size 201989
diff --git a/test/tatoeba-test.epo-vie.tsv b/test/tatoeba-test.epo-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..379a9fc6fa01e32d06111dd93dbf82765b91631a
--- /dev/null
+++ b/test/tatoeba-test.epo-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74ff21eca332dff3da7b9b2d730f04b0379957989bbdd948b83d5169c0721827
+size 173445
diff --git a/test/tatoeba-test.epo-vol.tsv b/test/tatoeba-test.epo-vol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c730c7dffa9c5eee6d28bc8d3cb509ca65a65eb5
--- /dev/null
+++ b/test/tatoeba-test.epo-vol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03839183ffd0037b07c3f3461be4866a8c4fa56b6b8eddacbee06125ef937558
+size 55056
diff --git a/test/tatoeba-test.epo-yid.tsv b/test/tatoeba-test.epo-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8c86b0f2d49ef42d1314a9ec24c025c90ee3efae
--- /dev/null
+++ b/test/tatoeba-test.epo-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:444c7350b3a0c67152e170c7771f496b98fbac29b68253156e6d03c2535840f2
+size 94278
diff --git a/test/tatoeba-test.epo-zho.tsv b/test/tatoeba-test.epo-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d771e79a4b0cf4051dc2c89dc9b3fcb8f0db4a68
--- /dev/null
+++ b/test/tatoeba-test.epo-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6ce0acc92e97cdfd21fa458de64924cbf894f2146388d43d0d65c2d7ab90590
+size 172567
diff --git a/test/tatoeba-test.est-rus.tsv b/test/tatoeba-test.est-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b64f09e5133c9dc787378047a3f457c354dae7fe
--- /dev/null
+++ b/test/tatoeba-test.est-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c709cbd93194bb394a0f15bbab0b034f4d98a991c8ba679bf8cd417d6c0915fa
+size 52902
diff --git a/test/tatoeba-test.eus-jpn.tsv b/test/tatoeba-test.eus-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d1380c3acce64177137caa3b66af280179e7b1bf
--- /dev/null
+++ b/test/tatoeba-test.eus-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a08561632426bd569bc43f8182f46587142ccb3a99254b6d1808265c77c7e220
+size 14683
diff --git a/test/tatoeba-test.eus-rus.tsv b/test/tatoeba-test.eus-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9e12a545ccc190f41971ffc37c59b16b762b9a61
--- /dev/null
+++ b/test/tatoeba-test.eus-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef4cdab585cf1a51e5a4cecea17333130169087f4e80e722176be79c3fdfcb7c
+size 35444
diff --git a/test/tatoeba-test.eus-spa.tsv b/test/tatoeba-test.eus-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f44024ff8080cb9032b74cf179f3f9a158b1509b
--- /dev/null
+++ b/test/tatoeba-test.eus-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac7bdc4085aa23e221717c629c82885d2a2f5708be600394e9a1b0164fc3aeac
+size 134389
diff --git a/test/tatoeba-test.fas-fra.tsv b/test/tatoeba-test.fas-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2b1e4799d5b82847da68457de321701669eb10bc
--- /dev/null
+++ b/test/tatoeba-test.fas-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2060df4b2c9f6b97427539a1186c142895efa0c14d1d2961d95dac1e08a7cb8
+size 45471
diff --git a/test/tatoeba-test.fin-fin.tsv b/test/tatoeba-test.fin-fin.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..04032103923834b606fcaa7df481b855a82c650b
--- /dev/null
+++ b/test/tatoeba-test.fin-fin.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7f7bab766adfe5da7065e28303fd37d0389e85ca9cd458a1f2d1c2d4e77438e
+size 66612
diff --git a/test/tatoeba-test.fin-fkv.tsv b/test/tatoeba-test.fin-fkv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4d1558a4a5ddc97ce7ccd50a6758f4c84fe5c76f
--- /dev/null
+++ b/test/tatoeba-test.fin-fkv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:758faffecb9835296e31730ab9b324952761e509d35705bde9b0710546989ca6
+size 32367
diff --git a/test/tatoeba-test.fin-fra.tsv b/test/tatoeba-test.fin-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1b9db31837e61b7a1a7ad550775a60b2ab62dd1d
--- /dev/null
+++ b/test/tatoeba-test.fin-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5daa4c61374ba4737cbef05a17f6f4acafdd704d00b471bf9e0169174519b5a
+size 131752
diff --git a/test/tatoeba-test.fin-heb.tsv b/test/tatoeba-test.fin-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d276ed8f5bedcdd3a54f80ec85d8df1fd131dceb
--- /dev/null
+++ b/test/tatoeba-test.fin-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca76953d3436b5c8c14ae11ff6a553428b95fcfc703fb38ccb38dcf610e709bc
+size 19871
diff --git a/test/tatoeba-test.fin-hun.tsv b/test/tatoeba-test.fin-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5b15eac5658665e3ea13fda268381b7c4c6f25d4
--- /dev/null
+++ b/test/tatoeba-test.fin-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58e7f914241918fea231d104b41cd7fccef6e86ad9b057fc02a70e01900fb0c2
+size 80998
diff --git a/test/tatoeba-test.fin-ita.tsv b/test/tatoeba-test.fin-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..104a6e40656caecc2bafcf079d2d7198b31d908e
--- /dev/null
+++ b/test/tatoeba-test.fin-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:422ea35e74c51045f286bdbd246c32ab12a414f30a86605a6948c62917beacd3
+size 72837
diff --git a/test/tatoeba-test.fin-jpn.tsv b/test/tatoeba-test.fin-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a9e87fe4f305ca3b23db8c5f4219c17e18d4f9ff
--- /dev/null
+++ b/test/tatoeba-test.fin-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e57eb306373aaf4dfd05851b5ba698743ce4ba8380d4a37307f695cc2e50f211
+size 938580
diff --git a/test/tatoeba-test.fin-jpn_Hani.tsv b/test/tatoeba-test.fin-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c0807ac7d81872f9e4eebe836bbb2ff63ab85f30
--- /dev/null
+++ b/test/tatoeba-test.fin-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03c7f8b8f553717d6515074801fccd6922785abcdbc0f442e592023d70fe59c9
+size 74484
diff --git a/test/tatoeba-test.fin-jpn_Hira.tsv b/test/tatoeba-test.fin-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ec54ea634adb44c270a074ed6eaa0cc15b2e86a5
--- /dev/null
+++ b/test/tatoeba-test.fin-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2999e621c5422981dd6385df78ac1ffe0dd6f0df5de84633e5bcb53fed756321
+size 841840
diff --git a/test/tatoeba-test.fin-jpn_Kana.tsv b/test/tatoeba-test.fin-jpn_Kana.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..86cd9efed89070d0fb968cba1063a41ee62332fc
--- /dev/null
+++ b/test/tatoeba-test.fin-jpn_Kana.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53d7611d51914820467297f557822ad1c2fd3203de9e9627a6b3bc50211039c6
+size 20953
diff --git a/test/tatoeba-test.fin-kor.tsv b/test/tatoeba-test.fin-kor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fcdfbddb05d472308ad65c02ed4c8fbc132443ca
--- /dev/null
+++ b/test/tatoeba-test.fin-kor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b81667148e5608e32a316fb5b4ee0a3f1e5641b14686d56014d8b91cdc8b236
+size 33898
diff --git a/test/tatoeba-test.fin-kor_Hang.tsv b/test/tatoeba-test.fin-kor_Hang.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7445b4a3a9e52175e27ebdce0624ef9e3e380b54
--- /dev/null
+++ b/test/tatoeba-test.fin-kor_Hang.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f87846c6486a81d7da9c2bc563830687ad9f83b2e9609b16c8615f722cc56c9a
+size 32896
diff --git a/test/tatoeba-test.fin-kur.tsv b/test/tatoeba-test.fin-kur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ad344d654e65cb5f8f68dc974ef82525d0ee1faa
--- /dev/null
+++ b/test/tatoeba-test.fin-kur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3eb862b960f3861dd25b06bc63fde9c757d413ed70b4f3b6babc5f8e6882132a
+size 20711
diff --git a/test/tatoeba-test.fin-lat.tsv b/test/tatoeba-test.fin-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3a3f6ba5c79fa42343bd64193a03e05bdf7a70c9
--- /dev/null
+++ b/test/tatoeba-test.fin-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5e8282010c6b8d33bbad423654fbbf6448c4206e328bf7f258ae026df0aa079
+size 18696
diff --git a/test/tatoeba-test.fin-nld.tsv b/test/tatoeba-test.fin-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..83a457b812573ad5789b7c2013060b36c0ac1497
--- /dev/null
+++ b/test/tatoeba-test.fin-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:375e5b3df438e90d76648b1c763e21c1da76a84c798a142039b70901dcea3c67
+size 21320
diff --git a/test/tatoeba-test.fin-nno.tsv b/test/tatoeba-test.fin-nno.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..24ecca60ccd23fdfa7c317f5316f940c8c0a9100
--- /dev/null
+++ b/test/tatoeba-test.fin-nno.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2126988d0e357b8f9d4e278c7e51a964879955aa50ca670985d3028221875ff
+size 50880
diff --git a/test/tatoeba-test.fin-nob.tsv b/test/tatoeba-test.fin-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4255994125ebf3674adcc010e02a592a526a858e
--- /dev/null
+++ b/test/tatoeba-test.fin-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfcb611c2f42afcb2b67b2764cfbd55c7d5e529a8d97732ea7980618694ae10f
+size 109733
diff --git a/test/tatoeba-test.fin-nor.tsv b/test/tatoeba-test.fin-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..22951f260f386938d359f4a6617a85b21a6666af
--- /dev/null
+++ b/test/tatoeba-test.fin-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:379addf806164ebe3c197b8c22f3f87644529faac613221a8cbc75ffa654a6db
+size 160613
diff --git a/test/tatoeba-test.fin-pol.tsv b/test/tatoeba-test.fin-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..253d21849d109cadec27dd149510d474871bf37c
--- /dev/null
+++ b/test/tatoeba-test.fin-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc4b7a4ee0ece0b7ea60c36d80f257f0d16848b3a5c9c570192b6bd0332a98f4
+size 43252
diff --git a/test/tatoeba-test.fin-por.tsv b/test/tatoeba-test.fin-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9fcf578061dd5bb06e3b773fe993f7af90a15de5
--- /dev/null
+++ b/test/tatoeba-test.fin-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd533bda1efc6d4f63fd1d46ba64eeb4d311f9eb3a77b6924de34c1e87f81ad2
+size 31930
diff --git a/test/tatoeba-test.fin-rus.tsv b/test/tatoeba-test.fin-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..445e2c2b885b6d9b6907a0883fe6c5f4e5d0153a
--- /dev/null
+++ b/test/tatoeba-test.fin-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be85c73dbb5d00647c8ac9b95453ca753cb24b653b0b70ee79494e52fe5ef92f
+size 323616
diff --git a/test/tatoeba-test.fin-spa.tsv b/test/tatoeba-test.fin-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bdc26ec3732c73683a520a611e82692280ad9bec
--- /dev/null
+++ b/test/tatoeba-test.fin-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b4a71f30740d6fb8e7c78bbb4a74395359e451eedc0db64dc50746b9c167611
+size 186328
diff --git a/test/tatoeba-test.fin-swe.tsv b/test/tatoeba-test.fin-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d0a9c8b229cbab53dc0957ce016971e325226124
--- /dev/null
+++ b/test/tatoeba-test.fin-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b364e10e2fadd84536f63a279676d88f06b9d6177621431b71c3dc9c773f988
+size 209037
diff --git a/test/tatoeba-test.fin-tur.tsv b/test/tatoeba-test.fin-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..77a784432e6e94c8513214cd99c855d85888e4d1
--- /dev/null
+++ b/test/tatoeba-test.fin-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:902619b9ba310128ee99c0608564b48f863901a216bfad85587dc8dc47e48226
+size 140929
diff --git a/test/tatoeba-test.fin-zho.tsv b/test/tatoeba-test.fin-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e205275c3e160ebbfa87584065c2b6b41370b4a2
--- /dev/null
+++ b/test/tatoeba-test.fin-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7197f8b2245b77c055d4ad0dae0c21c305334ab93e55613a838c4537849959dd
+size 24985
diff --git a/test/tatoeba-test.fra-cmn_Hans.tsv b/test/tatoeba-test.fra-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7c0e17e15de90b9c15b4743d0cfcb9630ba68b72
--- /dev/null
+++ b/test/tatoeba-test.fra-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60876ea4acd130bb81e7de3346f4398d18515586329f3196af3fceb2feb4cfa9
+size 632952
diff --git a/test/tatoeba-test.fra-cmn_Hant.tsv b/test/tatoeba-test.fra-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..532f81e348c595a50a548cca1370c59ae2871fef
--- /dev/null
+++ b/test/tatoeba-test.fra-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff8cc996f113ee2021913456a3d24f39cd5fba984ae3e440953874441842c131
+size 174852
diff --git a/test/tatoeba-test.fra-fra.tsv b/test/tatoeba-test.fra-fra.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c04fee18c86dd34348088f7e6d0de7ee84451fc0
--- /dev/null
+++ b/test/tatoeba-test.fra-fra.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a03349ac6c721a652ec1806afbf42f194de681e83f0c03d5648bf794c12ed12c
+size 84132
diff --git a/test/tatoeba-test.fra-gcf.tsv b/test/tatoeba-test.fra-gcf.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2233bea0379eec0c16aecd8da0fbfcb524f2e87b
--- /dev/null
+++ b/test/tatoeba-test.fra-gcf.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f45d7b8711c5cba8accff6f58084f3031c47cd054b77139655a28133d26e0f64
+size 69687
diff --git a/test/tatoeba-test.fra-hbs.tsv b/test/tatoeba-test.fra-hbs.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d7dc8b25ec7afb4338af755a93727578e7ff32bc
--- /dev/null
+++ b/test/tatoeba-test.fra-hbs.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa6464a5e95ce907ae29a07e275bfc7d11060ac153e95f6e982559b0e8036da9
+size 37837
diff --git a/test/tatoeba-test.fra-heb.tsv b/test/tatoeba-test.fra-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9cb3a6216b0dc9600a8d7fe8786062b9918d1ae5
--- /dev/null
+++ b/test/tatoeba-test.fra-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:903e5d9b2e75827bb8ccb7c25da639ec1091b88c6d0231a10176dffa80ca976e
+size 311865
diff --git a/test/tatoeba-test.fra-hrv.tsv b/test/tatoeba-test.fra-hrv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9bbe40e797f9161b852f9934aef27ddbfe62a6e0
--- /dev/null
+++ b/test/tatoeba-test.fra-hrv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85a5effbb5d64786715a6856519335862f83efabaae9485e6a39c9d44cf451c2
+size 19608
diff --git a/test/tatoeba-test.fra-hun.tsv b/test/tatoeba-test.fra-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1a5122842529e06607d85093c58f90572801fae8
--- /dev/null
+++ b/test/tatoeba-test.fra-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c04ecc99d4123d415c243af09ba491ca914175eaf70c5f7b9ce3e97d662b2466
+size 176603
diff --git a/test/tatoeba-test.fra-ido.tsv b/test/tatoeba-test.fra-ido.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4c44de8e4ba9056a76b622cf8ec1b0d94fe80301
--- /dev/null
+++ b/test/tatoeba-test.fra-ido.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bbd7169b78ca94b62ef068cb5deab4095ac48e47dc974a555f3f67bc25962e2
+size 20748
diff --git a/test/tatoeba-test.fra-ile.tsv b/test/tatoeba-test.fra-ile.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..75a14316d08942f2cec2ffc0cb5f016bf5614e48
--- /dev/null
+++ b/test/tatoeba-test.fra-ile.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36fcd37c70ee98bd35ec4820066ba18522be5a439ec4ea273b43012c9e21cf2d
+size 26879
diff --git a/test/tatoeba-test.fra-ina.tsv b/test/tatoeba-test.fra-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a735b1e78766d403e8b351a1948535a8d4db18fd
--- /dev/null
+++ b/test/tatoeba-test.fra-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8bacb545a8458d294da5bd497a2d918912e143e5c821aeb8f70962549ba40609
+size 105288
diff --git a/test/tatoeba-test.fra-ind.tsv b/test/tatoeba-test.fra-ind.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cf32bdea3360746b8a1eb0050ae75874b80ff4cd
--- /dev/null
+++ b/test/tatoeba-test.fra-ind.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f7fd789768d43b7872dbac31cefe7f0e130475334c186bb5e67f8eeb01575ca
+size 75212
diff --git a/test/tatoeba-test.fra-ita.tsv b/test/tatoeba-test.fra-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0d012fcf23f29fdfd8cb5b0265f868bdeac4caaf
--- /dev/null
+++ b/test/tatoeba-test.fra-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f8c9c40e63adba6e05d1b05e6ad724ef4bdd3e22ca133c73be4dd5a5b6cc039
+size 718494
diff --git a/test/tatoeba-test.fra-jbo.tsv b/test/tatoeba-test.fra-jbo.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..87ab6764ff3e82cc3906594a337dc9ff1117156c
--- /dev/null
+++ b/test/tatoeba-test.fra-jbo.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc0379e91cf7f8a802bef1e41898abdb1896ae65f2f8f8e7ec937c2ff29e6ba9
+size 94434
diff --git a/test/tatoeba-test.fra-jpn.tsv b/test/tatoeba-test.fra-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2ad07b7188b3d4700f4d0cd047653f31b4d381d3
--- /dev/null
+++ b/test/tatoeba-test.fra-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fa00570047daff0aa4ed05235637349945a6362a7914b22c0f93267bd7d1e95
+size 1102814
diff --git a/test/tatoeba-test.fra-jpn_Hani.tsv b/test/tatoeba-test.fra-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9db77ae5cb944f1b2b7b832616809b1c88a5a177
--- /dev/null
+++ b/test/tatoeba-test.fra-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5507b085fa05ffb5b78bdb9b071b4f763eec520d24cca51be2b4867994d0c33a
+size 116124
diff --git a/test/tatoeba-test.fra-jpn_Hira.tsv b/test/tatoeba-test.fra-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3c94c24601290109147806b8792611b6c0f64ed1
--- /dev/null
+++ b/test/tatoeba-test.fra-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9120ca1b63ee72af73ec923d79ea99da702c6dc4ea4aadfb6ef03943476fc5d
+size 965485
diff --git a/test/tatoeba-test.fra-kab.tsv b/test/tatoeba-test.fra-kab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cadeaaa206444b3997bee4fe303e085c2845d2cd
--- /dev/null
+++ b/test/tatoeba-test.fra-kab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2ec335e2b386fac928991583e27a674a0ca6d345f9d82062405621f05c43a4e
+size 1051694
diff --git a/test/tatoeba-test.fra-kor.tsv b/test/tatoeba-test.fra-kor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0d5f4bb1c427b455e6b1a7a10956dd88842b587b
--- /dev/null
+++ b/test/tatoeba-test.fra-kor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63ca05e4b74a65bd4dcdaa185159a84a4fe93e01d50ee83cb505beda2d3678a7
+size 32027
diff --git a/test/tatoeba-test.fra-kor_Hang.tsv b/test/tatoeba-test.fra-kor_Hang.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a556cb09a77d0f90ca2d9beb059869f4701c731e
--- /dev/null
+++ b/test/tatoeba-test.fra-kor_Hang.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:582ddd4786db0b94f3a0f98e2586fe77bbcff04a4fb346a67c8c0c5facf523d0
+size 31711
diff --git a/test/tatoeba-test.fra-lat.tsv b/test/tatoeba-test.fra-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..464425e46ee4529ffb3dd849b0ddcafff814d2b7
--- /dev/null
+++ b/test/tatoeba-test.fra-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f61569562feb13514be307e638ed3ef4159370fafec0b9743574cf3355123dc5
+size 364086
diff --git a/test/tatoeba-test.fra-lfn.tsv b/test/tatoeba-test.fra-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8ce786a5310d2968748978df403edd7030c0e38e
--- /dev/null
+++ b/test/tatoeba-test.fra-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1314e5137f14c106faf6db85289ed8c7f6a18efac3683482a03103c4e0a90f4
+size 55690
diff --git a/test/tatoeba-test.fra-lfn_Latn.tsv b/test/tatoeba-test.fra-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..96e60ccf942d1fdd49de5a6b02487c1d1c2899ed
--- /dev/null
+++ b/test/tatoeba-test.fra-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b060de2406f7d938475dfa8223cc973265718ed94a18344818c0b4d7089fc0b
+size 32585
diff --git a/test/tatoeba-test.fra-msa.tsv b/test/tatoeba-test.fra-msa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ec238a791dc64000a995cfda1a40470c620a9157
--- /dev/null
+++ b/test/tatoeba-test.fra-msa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cb97e42cf1f548982deebbdaf00be07064da71d02edfeb63d94aa23c97c682d
+size 85373
diff --git a/test/tatoeba-test.fra-nds.tsv b/test/tatoeba-test.fra-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f34826205636e5fca0572546338c7bec81676493
--- /dev/null
+++ b/test/tatoeba-test.fra-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31d7df9a6b5522ab19bc1d06e23bb79aea43ebefe12819bae7bf596a1c3fa310
+size 59528
diff --git a/test/tatoeba-test.fra-nld.tsv b/test/tatoeba-test.fra-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2b89f03cb03684f3226a14bc23eadf8a6fe23443
--- /dev/null
+++ b/test/tatoeba-test.fra-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e5eec627a24293b7c5855e8c661a12c2d816bbc38e18f6866aa4426c6645aa2
+size 894048
diff --git a/test/tatoeba-test.fra-nob.tsv b/test/tatoeba-test.fra-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c413f5e6825f1ec2526e3da50f3a3bf583d5b5e1
--- /dev/null
+++ b/test/tatoeba-test.fra-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53d69084a82db9efa5e92d9519649eb5f3e81a9ec51afbfe21cfc0a686084d04
+size 24303
diff --git a/test/tatoeba-test.fra-nor.tsv b/test/tatoeba-test.fra-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e22edd2a6f58f8b636bf0a5828e162d664d00345
--- /dev/null
+++ b/test/tatoeba-test.fra-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6c140e7b5586cfc636a7bcaa348b6a060a3ea25ff1cbfaba810f01779f65cc1
+size 33674
diff --git a/test/tatoeba-test.fra-oci.tsv b/test/tatoeba-test.fra-oci.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..998dcadff9d3649279bd2fe341087d0d0d391d6f
--- /dev/null
+++ b/test/tatoeba-test.fra-oci.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1bbc3063e9c2cff0dc9276de4d33dea2dfb23674f9fb87bc6675dda8e87b8dd
+size 68132
diff --git a/test/tatoeba-test.fra-pcd.tsv b/test/tatoeba-test.fra-pcd.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ae46a5f36d845ac3678ff942b8bee1efc9e890e1
--- /dev/null
+++ b/test/tatoeba-test.fra-pcd.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:699083287ee475908a384ab98fad5cd83f9645bb3803ad4020d94bac60794c67
+size 18060
diff --git a/test/tatoeba-test.fra-pol.tsv b/test/tatoeba-test.fra-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6040c28ba0ea2739192d8b45ea6e33343b88bcb9
--- /dev/null
+++ b/test/tatoeba-test.fra-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c727758f7a41bc5d818e49447690d7173d60524fa1d314504019cab7cf42856
+size 256127
diff --git a/test/tatoeba-test.fra-por.tsv b/test/tatoeba-test.fra-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e0ed1e5077434235bb44f45cba878126f7781bc3
--- /dev/null
+++ b/test/tatoeba-test.fra-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bf71bb7cce9fca7c04a84d26758495b2253631d2c3e092f1be0c338e55951e1
+size 858332
diff --git a/test/tatoeba-test.fra-ron.tsv b/test/tatoeba-test.fra-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0499c2652f07787a9cdb9210d5c4ad2d5f9be055
--- /dev/null
+++ b/test/tatoeba-test.fra-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4976be7eb3982736902eec2e204a175c63aa86ffdd89504d1671c65ce27b3f1d
+size 140669
diff --git a/test/tatoeba-test.fra-run.tsv b/test/tatoeba-test.fra-run.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ad91cec8cbacc3d995b086ae5e2d3de031ee7b26
--- /dev/null
+++ b/test/tatoeba-test.fra-run.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84e369a7be58ad20071248b22f84b881f9277490fa3320d0a9b38e7c290aeb34
+size 72697
diff --git a/test/tatoeba-test.fra-rus.tsv b/test/tatoeba-test.fra-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6fab3df4adeeed6ccc460d8fd9681805011f7b82
--- /dev/null
+++ b/test/tatoeba-test.fra-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dbfb5551b2f17f1e855aa4bc46115ae4c5577c10c17d52bf317f1673312e48f2
+size 1073575
diff --git a/test/tatoeba-test.fra-slv.tsv b/test/tatoeba-test.fra-slv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3ae4a87eebccbc47f47d75d40aa66fb1bb4429a9
--- /dev/null
+++ b/test/tatoeba-test.fra-slv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08e9128db00d42895a8fe1f8b4249e3e0e21e897c8507d78cea72a5104562816
+size 38259
diff --git a/test/tatoeba-test.fra-spa.tsv b/test/tatoeba-test.fra-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1705ef95f23dc02c0fcff1139c3b407424647ab7
--- /dev/null
+++ b/test/tatoeba-test.fra-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3ae44529915846db47267d74cf517b802dd57ad7d7ca68b2ae94a4cd1029c3a
+size 883143
diff --git a/test/tatoeba-test.fra-swe.tsv b/test/tatoeba-test.fra-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..214ac7509fbd0e3a3cd1320d62bbd9bef6feaab7
--- /dev/null
+++ b/test/tatoeba-test.fra-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8adec59ee73cd5f9542c62e7e1090abd6d35736e77892b7f9ca1d20fcb5899a8
+size 101365
diff --git a/test/tatoeba-test.fra-tat.tsv b/test/tatoeba-test.fra-tat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b7d29400a4be6543d340a89e831d180489f42f4f
--- /dev/null
+++ b/test/tatoeba-test.fra-tat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02c7200e7cab52f1a5942ea75fbab605ed05de5033e9f38726dc7ce53d50337c
+size 25778
diff --git a/test/tatoeba-test.fra-tgl.tsv b/test/tatoeba-test.fra-tgl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a04333eabca2f75996c81346172e1c994d6dc0d2
--- /dev/null
+++ b/test/tatoeba-test.fra-tgl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c07a8138e731f965acc621826132804e752ba4a028b217decefff3531f08ca2
+size 69627
diff --git a/test/tatoeba-test.fra-tlh.tsv b/test/tatoeba-test.fra-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a5cce1eb1dcac3db3df736e740f2ebb05ca60cdd
--- /dev/null
+++ b/test/tatoeba-test.fra-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:141de91fcdfa8f5c79e070d3d681f61765f1cb8dc3344d32b31cd7ac62c1ca50
+size 41550
diff --git a/test/tatoeba-test.fra-tlh_Latn.tsv b/test/tatoeba-test.fra-tlh_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2ab3666f9c3e5791ff6e085d0565730eb1312902
--- /dev/null
+++ b/test/tatoeba-test.fra-tlh_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7582ce47f204bc828102497dc41ad5103b079dd5cdde57e2b5b625317ee74cf3
+size 41512
diff --git a/test/tatoeba-test.fra-toki.tsv b/test/tatoeba-test.fra-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..65aba3f81fc4e561ec31a456b833d529102923e6
--- /dev/null
+++ b/test/tatoeba-test.fra-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7332980b2d54d71a5e3aed058f758f1fcc51dcb29558ba4897fdd5935169892e
+size 48138
diff --git a/test/tatoeba-test.fra-toki_Latn.tsv b/test/tatoeba-test.fra-toki_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6045746e699f2deecb5c6dc9756b4711ea828af3
--- /dev/null
+++ b/test/tatoeba-test.fra-toki_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35f46fede0287a2dceda0749492e3a2f396eb51a84b6f2bc52737a42549faf9d
+size 48117
diff --git a/test/tatoeba-test.fra-tur.tsv b/test/tatoeba-test.fra-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9e966fc4998b30bc86919796904ee6013c0f467e
--- /dev/null
+++ b/test/tatoeba-test.fra-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d8cfcd3afe5e0c16d85aac2bcd454927c782f80559a83b8d33aa546d909b864
+size 208919
diff --git a/test/tatoeba-test.fra-uig.tsv b/test/tatoeba-test.fra-uig.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..79e46405efaab4bf8c92b1a59bb77056434afca9
--- /dev/null
+++ b/test/tatoeba-test.fra-uig.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cac223581ca9a6b2e81f933a95adda32f734cd5d250dae25aba9d0a6a7d91d0a
+size 66867
diff --git a/test/tatoeba-test.fra-uig_Arab.tsv b/test/tatoeba-test.fra-uig_Arab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9061251640ff7a0233fc362dd531435d8f206fa5
--- /dev/null
+++ b/test/tatoeba-test.fra-uig_Arab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbaeeeb377d962e02e874abaa2bbce7092da487021c5b58e8dfb2b3802946b6a
+size 66689
diff --git a/test/tatoeba-test.fra-ukr.tsv b/test/tatoeba-test.fra-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7e3c5b2b39d0694a83fea5ecb00a9a7be217bc07
--- /dev/null
+++ b/test/tatoeba-test.fra-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f43108958f506921afa52b4cfe4f3a925510de30a8a1750ffc11d764768a39d
+size 836473
diff --git a/test/tatoeba-test.fra-vie.tsv b/test/tatoeba-test.fra-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8270cbbbf54befa2c241dfffa6fb263d25f32efe
--- /dev/null
+++ b/test/tatoeba-test.fra-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86c9ce29a679ed8a237d1d01dc62e729408391e0395edca61837e79134145c73
+size 136035
diff --git a/test/tatoeba-test.fra-wuu.tsv b/test/tatoeba-test.fra-wuu.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..89e4807792ee305d54e1701e9b6c2e6d700f261f
--- /dev/null
+++ b/test/tatoeba-test.fra-wuu.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1309d7a1c478a3e7ffd4e641a2c7724b735a59be77143cefef9b9197f0fdb8f1
+size 65808
diff --git a/test/tatoeba-test.fra-yid.tsv b/test/tatoeba-test.fra-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4ebec1b469efed7cd8f6e2ba31406a10ccc3e59a
--- /dev/null
+++ b/test/tatoeba-test.fra-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a02c0445bfa0773e8f75232e3bb56240fd725a3e7b7cb55d7a335e8260af951e
+size 35558
diff --git a/test/tatoeba-test.fra-zho.tsv b/test/tatoeba-test.fra-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..70427b2cdc754f14756b29f356d84eef46ee17c6
--- /dev/null
+++ b/test/tatoeba-test.fra-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60ac0c5c6c0dff4d27429dcb2df8cdca370f4d38b580625869b4c4ea938bc947
+size 896410
diff --git a/test/tatoeba-test.fry-nld.tsv b/test/tatoeba-test.fry-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cb48ec2c5c5c2eda6f4554e3a47ac24584afaa19
--- /dev/null
+++ b/test/tatoeba-test.fry-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73b83362ee2d8a20305708130fbcb7e32527d06003bbb181efd3873e7f85bd89
+size 18557
diff --git a/test/tatoeba-test.gcf-gcf.tsv b/test/tatoeba-test.gcf-gcf.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0e5c270e455d4a9dab3c5d55e46ba944b66201ad
--- /dev/null
+++ b/test/tatoeba-test.gcf-gcf.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c1183a5c8dc9727a19bffd6d0f3ca96726792d5702e77046a97bd4d6c3d55a0
+size 14350
diff --git a/test/tatoeba-test.gla-spa.tsv b/test/tatoeba-test.gla-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5d558e95563e0b58571cf0d21a64bf011a0258d9
--- /dev/null
+++ b/test/tatoeba-test.gla-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:172f8f35f1ba38942eed73a59a28c154b9a2429cecc5ca856562cf34be763d84
+size 19141
diff --git a/test/tatoeba-test.glg-por.tsv b/test/tatoeba-test.glg-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..78db41136d6ab522e0e306794967fc3d7f2250aa
--- /dev/null
+++ b/test/tatoeba-test.glg-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e88b7f63142a11a4b8babbfa33564b911ba1ad70256e0fd82f4f1e351362610
+size 33423
diff --git a/test/tatoeba-test.glg-spa.tsv b/test/tatoeba-test.glg-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c85cbb3d078158e7d7dd4548acb50ab72f8ba225
--- /dev/null
+++ b/test/tatoeba-test.glg-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8ba38757d5ff9bc2c6fb6a1b48ae03b5fb7c53f34df6e561d60f5e2d2ee70e3
+size 183901
diff --git a/test/tatoeba-test.gos-nld.tsv b/test/tatoeba-test.gos-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..acee645255786760735b1eb5b56a98bca402e1c2
--- /dev/null
+++ b/test/tatoeba-test.gos-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87fc153573c69122a9a6354484eb300455610f29839330a21dc7414225caecaf
+size 100409
diff --git a/test/tatoeba-test.grn-por.tsv b/test/tatoeba-test.grn-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..82c3c316b05747c9b0c36f6e2abee2a9cb499ec8
--- /dev/null
+++ b/test/tatoeba-test.grn-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2aadaa888abb69f23e2976cee6b82054e16a80e34d5cdd5a616bb5b14483c45b
+size 17641
diff --git a/test/tatoeba-test.grn-spa.tsv b/test/tatoeba-test.grn-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2cd68536b6f6006ad2e13207baf02503457413bd
--- /dev/null
+++ b/test/tatoeba-test.grn-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef5f2b1c2b21f697df24d1d1bba1cc28c068bf9c338c8fe5ad3b0ba1c9d939ce
+size 45645
diff --git a/test/tatoeba-test.hbs-ita.tsv b/test/tatoeba-test.hbs-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ae071784fc00806fbaa4543f7a7ba3d762b49dbc
--- /dev/null
+++ b/test/tatoeba-test.hbs-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32193ba0dbeafe2f4496ab693fdb948e807817a6d94252b0e2e6e16f5dcdd426
+size 37083
diff --git a/test/tatoeba-test.hbs-jpn.tsv b/test/tatoeba-test.hbs-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2343199e1ad20d415b31a0b8aaf371b82b4c5c29
--- /dev/null
+++ b/test/tatoeba-test.hbs-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f0c222382e65aeea5dc8b98a732de6c40702b0714fdd4606bc00dc32b46e233
+size 35763
diff --git a/test/tatoeba-test.hbs-nor.tsv b/test/tatoeba-test.hbs-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0f7aaaaf9f6db597ec9ab22e23545db9f933bdef
--- /dev/null
+++ b/test/tatoeba-test.hbs-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cae0c314c6761e581b99f6c96d15a3c503cbbb30dc9f1ee1f4ab05040ce6fefe
+size 481328
diff --git a/test/tatoeba-test.hbs-pol.tsv b/test/tatoeba-test.hbs-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dc3ac303eed4a767ecedde486b59b57d7013ed76
--- /dev/null
+++ b/test/tatoeba-test.hbs-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eac1a614cbce6ef85b5355096dc8c6029361a03127c644fbfafa3b3b61528d60
+size 33128
diff --git a/test/tatoeba-test.hbs-rus.tsv b/test/tatoeba-test.hbs-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f591f83840bbf561bb5c84ee172193561d4b3299
--- /dev/null
+++ b/test/tatoeba-test.hbs-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:932f24477924ed03a00332e5e2893d8f1e0babb93dfcce40e5f714d16a5899f6
+size 237071
diff --git a/test/tatoeba-test.hbs-spa.tsv b/test/tatoeba-test.hbs-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3272996c0ebd53fb3b07baa0b77aaa3a7c36b79b
--- /dev/null
+++ b/test/tatoeba-test.hbs-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9bb6068caa2a6236d15ff7e8a9be4467af52dabaee93196f5ba68ec91787208
+size 42872
diff --git a/test/tatoeba-test.hbs-ukr.tsv b/test/tatoeba-test.hbs-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c02cab72c50e68775f0a3baab331e449f23b2bdf
--- /dev/null
+++ b/test/tatoeba-test.hbs-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea45b01982cf4100bee86f7d4df9d99e221472525a1a7d63aad7bda3e9bac214
+size 76362
diff --git a/test/tatoeba-test.hbs-zho.tsv b/test/tatoeba-test.hbs-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dcec6e1b6eadfa7f7911872bc2fb3154940e7973
--- /dev/null
+++ b/test/tatoeba-test.hbs-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b71e94cdbd39c4936e613181397c96df6e516ed4026d8595a612f67c85bc96bd
+size 17387
diff --git a/test/tatoeba-test.heb-cmn_Hans.tsv b/test/tatoeba-test.heb-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..22f13f916127ee54bd5ac6091ac87678e7e93e84
--- /dev/null
+++ b/test/tatoeba-test.heb-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67c9d92395ae7488b09807d83b2f9c4339c4daf49a701603b4f26ec0b27a945d
+size 22756
diff --git a/test/tatoeba-test.heb-cmn_Hant.tsv b/test/tatoeba-test.heb-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..360e4f35c8994be06931256ad59d8fa068381e42
--- /dev/null
+++ b/test/tatoeba-test.heb-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cee81d3fa04fb84147795a8343aa8178466efb2c6087e57455e26dacb7e9d103
+size 23289
diff --git a/test/tatoeba-test.heb-heb.tsv b/test/tatoeba-test.heb-heb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4d245c01119a62e279c8f70e934cf7e41059300e
--- /dev/null
+++ b/test/tatoeba-test.heb-heb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:324d5747bc751b1d65fabee78ce439193f85ee9f959c3078c940dce984056148
+size 96379
diff --git a/test/tatoeba-test.heb-hun.tsv b/test/tatoeba-test.heb-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a1674b9f273c3e533e7bc3ce69f0085835a48b61
--- /dev/null
+++ b/test/tatoeba-test.heb-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:acc4170c14cbbd30ca59ea90f4681aca5d0086bbf885cf17a3f69b2592bd13fb
+size 32411
diff --git a/test/tatoeba-test.heb-ina.tsv b/test/tatoeba-test.heb-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0310d1dc674e3a3849d1d449ef100838710ba85d
--- /dev/null
+++ b/test/tatoeba-test.heb-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3b76874e15d0606f95f53cf41a08df35c9dfabe5d71e6a066fbf9f68273329e
+size 19332
diff --git a/test/tatoeba-test.heb-ita.tsv b/test/tatoeba-test.heb-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..aa9fefac7c4de0572928f27725e12b79958356d8
--- /dev/null
+++ b/test/tatoeba-test.heb-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a2c3243227e60e423a558bbedec700d634778302c4e310ffcc710c7ec7eed5f
+size 142984
diff --git a/test/tatoeba-test.heb-jpn.tsv b/test/tatoeba-test.heb-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e44cbd7a3d70fd658a8c666df0a6188f4df66cac
--- /dev/null
+++ b/test/tatoeba-test.heb-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c66b6c698fd7a1b5e98b28231ecf61545fcb72a6e68433d6332afc7059b34c5
+size 27471
diff --git a/test/tatoeba-test.heb-jpn_Hira.tsv b/test/tatoeba-test.heb-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dd4f3b597a2fa8ffa5d607328ac1e18e1a817cca
--- /dev/null
+++ b/test/tatoeba-test.heb-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d93c91e7c124ab67af681a631e94982bf885fb670575e5255c2b69fa1c24b00d
+size 23573
diff --git a/test/tatoeba-test.heb-lad.tsv b/test/tatoeba-test.heb-lad.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c87c1d5bdcaf96f7b7a9dfc4d60355a66e7c4b14
--- /dev/null
+++ b/test/tatoeba-test.heb-lad.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f73d3c98635d3836beea349bad5be2ca7b8ad749ae59613202db42e6076ae2b
+size 18893
diff --git a/test/tatoeba-test.heb-lat.tsv b/test/tatoeba-test.heb-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..99058624c45d8bd0fbfef730e84a26a62b039c5a
--- /dev/null
+++ b/test/tatoeba-test.heb-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a13b23d9f659b5cd2ec2d365920ca6d4920ec534c59aa3b1ae3f540a4fdc94f6
+size 23651
diff --git a/test/tatoeba-test.heb-lfn.tsv b/test/tatoeba-test.heb-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7b3f03effdac21b7afb57df0cdedb1654bb51228
--- /dev/null
+++ b/test/tatoeba-test.heb-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:affc69230bf2e9b10e70c93c44c76719d0baf9b87ce60527194c4e53d8d28223
+size 24022
diff --git a/test/tatoeba-test.heb-lfn_Latn.tsv b/test/tatoeba-test.heb-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c56a76f44fcfd46f6fa99cd78bd0228a687e92a7
--- /dev/null
+++ b/test/tatoeba-test.heb-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8951259677dc74ef9794a8046e6c664e14c9422e21e1203adc05df4931fa8301
+size 17299
diff --git a/test/tatoeba-test.heb-nld.tsv b/test/tatoeba-test.heb-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4cdd86d80d8eb731117c3841b1d89bff4ba96292
--- /dev/null
+++ b/test/tatoeba-test.heb-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25679dfd19e078b1c45a59206fc1fc9d0bd146645f8d7368182424a658cffc33
+size 236821
diff --git a/test/tatoeba-test.heb-pol.tsv b/test/tatoeba-test.heb-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..11d2fc38f9a3d784590b4e8837c247f619fecd31
--- /dev/null
+++ b/test/tatoeba-test.heb-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7d5a4ff07697e11664c5f9e8d781ea6ce8db0f98e71084cae045e4be6a1000f
+size 437410
diff --git a/test/tatoeba-test.heb-por.tsv b/test/tatoeba-test.heb-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cf6a1d9b0db8137ba179d5c4cdb8a5ae9462d6b9
--- /dev/null
+++ b/test/tatoeba-test.heb-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2ebb95f77d276edf80862d7d723d6f38c7082094c397b6dfdf3ad01922cc04a
+size 64157
diff --git a/test/tatoeba-test.heb-rus.tsv b/test/tatoeba-test.heb-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..981bd9a22f9d96a0317fc0ff03f0716e13485867
--- /dev/null
+++ b/test/tatoeba-test.heb-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7c1d48cc64f754887d40c76414c56c72e9b8d0413b512a01a88c29e34b048d5
+size 272864
diff --git a/test/tatoeba-test.heb-spa.tsv b/test/tatoeba-test.heb-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..833382f4a1bc331e8643118d8289679f0c976d40
--- /dev/null
+++ b/test/tatoeba-test.heb-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6eaedb19abb7b1a1878b50d05ff5f0946551f68a48cd779c20574e49ff3d1415
+size 174876
diff --git a/test/tatoeba-test.heb-tur.tsv b/test/tatoeba-test.heb-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..87acb62b0c44907fa7207150592d7f63db643f6f
--- /dev/null
+++ b/test/tatoeba-test.heb-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7028b59201b5f3cbe56475ca986705af735d6f57a94a71e3ebb3b3735d1fd685
+size 104779
diff --git a/test/tatoeba-test.heb-ukr.tsv b/test/tatoeba-test.heb-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0c4256260642efd179bbc3e6e6833d33a2482acf
--- /dev/null
+++ b/test/tatoeba-test.heb-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26285b5c3ae9d16f52087776b1d2a7ea6e0767620f97b595d56efc85f0c99e62
+size 83517
diff --git a/test/tatoeba-test.heb-yid.tsv b/test/tatoeba-test.heb-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3fb0a502777bc4500aa2d763628a33697ddd3857
--- /dev/null
+++ b/test/tatoeba-test.heb-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca054f3bb2c39f978bc066c56941525ab2b912ed38e5f18c7bd80003fb8db05f
+size 98653
diff --git a/test/tatoeba-test.heb-zho.tsv b/test/tatoeba-test.heb-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7a80d876fb82ffa8834db9e0cb38ed73f83e92ed
--- /dev/null
+++ b/test/tatoeba-test.heb-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9efe0842ad442988d4aec95d71c502bad118c75c9745f627b32908273a183da4
+size 52399
diff --git a/test/tatoeba-test.hin-urd.tsv b/test/tatoeba-test.hin-urd.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b684b95c01b39f9504b83d8b2d9df1daac07bc3e
--- /dev/null
+++ b/test/tatoeba-test.hin-urd.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e3b817cee5c2425a83a7d11f55eaded41dbc044efbaf8820450038873089370
+size 33691
diff --git a/test/tatoeba-test.hin-zho.tsv b/test/tatoeba-test.hin-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a7050711874d12d63d2c3d9397764824d2e608cb
--- /dev/null
+++ b/test/tatoeba-test.hin-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7864b74d1413f38dbf2c7d721f254ee21c7e9f7f2a5b8794188eb8947e1f4e3c
+size 44648
diff --git a/test/tatoeba-test.hrv-jpn_Hira.tsv b/test/tatoeba-test.hrv-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..52c8df15f925e946727b6ec7356c3e09885e595b
--- /dev/null
+++ b/test/tatoeba-test.hrv-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cc9aa27f11c838c0e68fee2e30b585fe0b0c5bcc922070db741d218fb6ec7ae
+size 24963
diff --git a/test/tatoeba-test.hrv-pol.tsv b/test/tatoeba-test.hrv-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..48ff7cb3a8557465f106fd99de359410c11d288f
--- /dev/null
+++ b/test/tatoeba-test.hrv-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d59801dca3a99e0187de3abeb71d3707eb9b0d32ffa6a3fd55a8b8adaed738c0
+size 19473
diff --git a/test/tatoeba-test.hrv-spa.tsv b/test/tatoeba-test.hrv-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..49a90a3a7c328adf44643386a160e7a42b3a9235
--- /dev/null
+++ b/test/tatoeba-test.hrv-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c44b2c6e0438b625a6d0b1114a4e574d8d9a50942093c3107c57dece436ac495
+size 16960
diff --git a/test/tatoeba-test.hrv-ukr.tsv b/test/tatoeba-test.hrv-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..039600dac494b0bd7ee8f6127a72344dea69020b
--- /dev/null
+++ b/test/tatoeba-test.hrv-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59013b176f3cb641acb17031fd25c600c502c2a42661a6669a82566fd681286a
+size 32312
diff --git a/test/tatoeba-test.hsb-slv.tsv b/test/tatoeba-test.hsb-slv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..51add1fb675064c23dcd94b0c3102c544cab5195
--- /dev/null
+++ b/test/tatoeba-test.hsb-slv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d4c6cb9552157ddf1ce848baa2efba1f35ccb5215118551506ec313fde0d93a
+size 35522
diff --git a/test/tatoeba-test.hun-cmn_Hans.tsv b/test/tatoeba-test.hun-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ec67c07eaf6efc22f2db4c656bfb21d8896b6f3b
--- /dev/null
+++ b/test/tatoeba-test.hun-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f266c51b859266a6c35edc66c6d55ab0e2d6ae1e773e45b7d33d209949f668e8
+size 19862
diff --git a/test/tatoeba-test.hun-hun.tsv b/test/tatoeba-test.hun-hun.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..045992feeb515d2fdc799ed94e0493b24808590c
--- /dev/null
+++ b/test/tatoeba-test.hun-hun.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9809caec12800fbabf5419fd55201a440dbcece432e0ea4c99aa9f845a70121b
+size 84532
diff --git a/test/tatoeba-test.hun-ita.tsv b/test/tatoeba-test.hun-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3c264b0306dd62269f89d5222416f5baa2ae7230
--- /dev/null
+++ b/test/tatoeba-test.hun-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f1a34b0686e0baf5f5135453328c781ea932a06477a9a6f3525f1653eaebb59
+size 329245
diff --git a/test/tatoeba-test.hun-jpn.tsv b/test/tatoeba-test.hun-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0dbd33b8be0f3aeb337c1b11277415b661ef59c3
--- /dev/null
+++ b/test/tatoeba-test.hun-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:855d37c4e5e8fda8c5785299a5ce3bf36b9601954cab95253c3dd30cbf9586ab
+size 282076
diff --git a/test/tatoeba-test.hun-jpn_Hani.tsv b/test/tatoeba-test.hun-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2f899c7eaa176775f0b15917051c2d49bf179f5a
--- /dev/null
+++ b/test/tatoeba-test.hun-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d330dccd9be7accce360f5daadf9be1a7c31a775c369405ea2cd6c36009a4162
+size 37322
diff --git a/test/tatoeba-test.hun-jpn_Hira.tsv b/test/tatoeba-test.hun-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bf2d740e925e0523370b28b105269ba1ff097e69
--- /dev/null
+++ b/test/tatoeba-test.hun-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:599639fb1122fb123139aca8e5f6e7c9c86224d625215986bbe0d5f950dc2d1d
+size 242531
diff --git a/test/tatoeba-test.hun-kor.tsv b/test/tatoeba-test.hun-kor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..66be804b5ea8e66ed429a1165131859b39d05911
--- /dev/null
+++ b/test/tatoeba-test.hun-kor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ffd3ed140071cdef5259532a3aaaac56d05b02ed7eb2740b0857174b9d08dcf
+size 22711
diff --git a/test/tatoeba-test.hun-kor_Hang.tsv b/test/tatoeba-test.hun-kor_Hang.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b58cd6715989e725cdf0345650ca887d040c4ccd
--- /dev/null
+++ b/test/tatoeba-test.hun-kor_Hang.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5215f16d7d10136abd745e150ab41d1d32b60defd5e03350aca550913c3c1342
+size 22473
diff --git a/test/tatoeba-test.hun-lat.tsv b/test/tatoeba-test.hun-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e24f3dea4965f0fe08b9b24943e321a9241a480c
--- /dev/null
+++ b/test/tatoeba-test.hun-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b061bc96b0ca194ded274aa8b772a78820a9a2bd556874909ba3ca115194b08c
+size 43106
diff --git a/test/tatoeba-test.hun-nld.tsv b/test/tatoeba-test.hun-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..05d8b8d06095145549d557a31adb57651ee558b0
--- /dev/null
+++ b/test/tatoeba-test.hun-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba13eba43ec173035d3b419427eea8ddd2657b2a8ec3b1d7e72bf1f203e15540
+size 113849
diff --git a/test/tatoeba-test.hun-pol.tsv b/test/tatoeba-test.hun-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..902f6fbd05a832020e8ac5059cc8c243c546617f
--- /dev/null
+++ b/test/tatoeba-test.hun-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0beb68fda8826eff06378633ea74c66d2d01ad37697b22e447b80fb2aa8dd49d
+size 140082
diff --git a/test/tatoeba-test.hun-por.tsv b/test/tatoeba-test.hun-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..05c7fbfe0e9af9c983344b132d7757513f3ea425
--- /dev/null
+++ b/test/tatoeba-test.hun-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:894665bda5ccf1ee53601b64870afbc2c9364acc6d0657ee05f9ca2f21abf420
+size 176114
diff --git a/test/tatoeba-test.hun-rus.tsv b/test/tatoeba-test.hun-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..81baa4ffb68e434b61b11c1b4b9ebbdc8f1b6a86
--- /dev/null
+++ b/test/tatoeba-test.hun-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51f2fe46163e63883e397555f0742b8dc91da10ebc3127def427017a024ee9fa
+size 252859
diff --git a/test/tatoeba-test.hun-spa.tsv b/test/tatoeba-test.hun-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..50cb20b1ff03dd217e7c6599e28b58dc45fe7723
--- /dev/null
+++ b/test/tatoeba-test.hun-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a6b13f78f4b1357a51fc100f5341cb370216466b7df66f5a4c3a1882b9373c6
+size 179662
diff --git a/test/tatoeba-test.hun-swe.tsv b/test/tatoeba-test.hun-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9be73881a7eda65f6909a21229c2b14f8a5f7e95
--- /dev/null
+++ b/test/tatoeba-test.hun-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5eabf35ccea353c4146bcfd8250f60b8acb00ffc34142c8f35aabbca9440aef1
+size 117937
diff --git a/test/tatoeba-test.hun-tur.tsv b/test/tatoeba-test.hun-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9dce56aa5af1af66c44e927d2d2e30bea2fae662
--- /dev/null
+++ b/test/tatoeba-test.hun-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e51fb182da8b9f68af2ac8826a7eed9bc08c27337e788cc16629cbffbcea10a
+size 63469
diff --git a/test/tatoeba-test.hun-ukr.tsv b/test/tatoeba-test.hun-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f5d993488ae5331f893df6ca2fd5092b09ca06b1
--- /dev/null
+++ b/test/tatoeba-test.hun-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc33c5603b90098ca68ca2863815ac80bb839ddd592c783477669b99827d8d5b
+size 37775
diff --git a/test/tatoeba-test.hun-zho.tsv b/test/tatoeba-test.hun-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..edf9395fbb82f278fa2ac39dd3eeed60c74b63a7
--- /dev/null
+++ b/test/tatoeba-test.hun-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:522142bb06cb052f571f64b0270995addc206a5fbe031d24cac8b300f04b7d31
+size 33218
diff --git a/test/tatoeba-test.hye-rus.tsv b/test/tatoeba-test.hye-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dff483bee4dc0e9a16c5302f1072b562fbe8a51f
--- /dev/null
+++ b/test/tatoeba-test.hye-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d6cefe32a2baf7f6d27505e245991e5c7a6fc1a48be12d3e36069a6ed7a7fcd
+size 29846
diff --git a/test/tatoeba-test.ido-ina.tsv b/test/tatoeba-test.ido-ina.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d2ec6097f194db195ee1a29b4212991b480edb36
--- /dev/null
+++ b/test/tatoeba-test.ido-ina.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:754d29a9e7d8ab32a6af8c38e51e1e08ff9f3b18962e7772393f989b669265bd
+size 20767
diff --git a/test/tatoeba-test.ido-ita.tsv b/test/tatoeba-test.ido-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..87efac5469793870aa5dff7a647f63c91cbd468f
--- /dev/null
+++ b/test/tatoeba-test.ido-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:035b1b3d024a4c8e03070acf8d7739d1058c3068a28581cf9ff269289264ffc1
+size 114205
diff --git a/test/tatoeba-test.ido-lfn.tsv b/test/tatoeba-test.ido-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8862980037d1189e62244c3b8bd803d37bc4a3f7
--- /dev/null
+++ b/test/tatoeba-test.ido-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83827eeee94acb5987efa7df4af85498cc6d21f2a8efb822be6502e15aa8847b
+size 23776
diff --git a/test/tatoeba-test.ido-spa.tsv b/test/tatoeba-test.ido-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b42a55b9556927ed4dac5b8b14067a50ff0c1f9f
--- /dev/null
+++ b/test/tatoeba-test.ido-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23863b0f6c38ea9037428032cd2217903f010cfa44140576b0ef98ee88c1a871
+size 37025
diff --git a/test/tatoeba-test.ido-yid.tsv b/test/tatoeba-test.ido-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8ad69ef58579937b9a36d08182a0089819954f5d
--- /dev/null
+++ b/test/tatoeba-test.ido-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66973463b8f0dff8fad3e01bea2814d35d78e33ad2a51b95b1bf255940f60e71
+size 37649
diff --git a/test/tatoeba-test.ido_Latn-lfn_Latn.tsv b/test/tatoeba-test.ido_Latn-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..355b8c59836bb82689b1fe9537fc9ecd4b5e4015
--- /dev/null
+++ b/test/tatoeba-test.ido_Latn-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a07f069bc718b025b09c65398b51740ab90624c83d6b975f04a600f2e017e4df
+size 20200
diff --git a/test/tatoeba-test.ina-ita.tsv b/test/tatoeba-test.ina-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..20f628969dcb73b0126caa6b8b2238e6d4b54f57
--- /dev/null
+++ b/test/tatoeba-test.ina-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b8e7bafd8c5bc54a4ed978e1d712cf1c7f412a9e1df07faa985399c3ffc7b2a
+size 80083
diff --git a/test/tatoeba-test.ina-lad.tsv b/test/tatoeba-test.ina-lad.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6617e4d46a5919e9e3a5e4a40e422d9acbf79e65
--- /dev/null
+++ b/test/tatoeba-test.ina-lad.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c060f8f060833697193aad36aecfa3ecf0358c4a17669a437799c685f8e2a2ec
+size 23470
diff --git a/test/tatoeba-test.ina-lat.tsv b/test/tatoeba-test.ina-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bf60280b42bb1b395cba800fe050112ad4455801
--- /dev/null
+++ b/test/tatoeba-test.ina-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65d3afc1367afce6a06726cd83106d38c6ed34fbbb30a43b2661b541bdbd6652
+size 72447
diff --git a/test/tatoeba-test.ina-lfn.tsv b/test/tatoeba-test.ina-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4e84e86e9457a15885685726599b466c4dd6ace7
--- /dev/null
+++ b/test/tatoeba-test.ina-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40659eeac476294e9cde602e5a3d154f2d346d76374a6dee63da79d4d7852c56
+size 55918
diff --git a/test/tatoeba-test.ina-nld.tsv b/test/tatoeba-test.ina-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..124da7f392f31fbea73a1b78d25bda485cad2e11
--- /dev/null
+++ b/test/tatoeba-test.ina-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c37c413463c354c1700def5d2c65f61bebfdc20d4dd4c02118cd9fe53b44058
+size 90977
diff --git a/test/tatoeba-test.ina-por.tsv b/test/tatoeba-test.ina-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9227628d2bddf83199015916c551529c3cd9ee80
--- /dev/null
+++ b/test/tatoeba-test.ina-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b519698483d8d5480d129a8e691f5b027e8974a91d1b79b4b00b3f4fc646402
+size 275684
diff --git a/test/tatoeba-test.ina-rus.tsv b/test/tatoeba-test.ina-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a9e6c0e3bc9f508c645e04999146494cd224dc2e
--- /dev/null
+++ b/test/tatoeba-test.ina-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:def4f4d072b9861e4196b07d4b7cf3151ae77edebafff777ba6f2b775a810f60
+size 175954
diff --git a/test/tatoeba-test.ina-spa.tsv b/test/tatoeba-test.ina-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..637614a1d074bd8f1812c9167f623c3162fcee20
--- /dev/null
+++ b/test/tatoeba-test.ina-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:574b82d019a723ecaa0655b5fea5b4e51e88c66959cea473e1974b172705f353
+size 165605
diff --git a/test/tatoeba-test.ina-tlh.tsv b/test/tatoeba-test.ina-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c6acee6994e6002cca0869479606436b0b4d9bf1
--- /dev/null
+++ b/test/tatoeba-test.ina-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4404cddde6dd334a19c19dc348d3b689f3f3d8cfd19f7c15e6abf7c62d1f5435
+size 12352
diff --git a/test/tatoeba-test.ina-tur.tsv b/test/tatoeba-test.ina-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..35e24ab3dba38ac842110cab0526a16d32f8317a
--- /dev/null
+++ b/test/tatoeba-test.ina-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5fcdea3a6642f65d42a0334c5aacb64279ca4b17af7e1db745ab3e61cbecd35
+size 20364
diff --git a/test/tatoeba-test.ina-yid.tsv b/test/tatoeba-test.ina-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..93b23ad3adf5198f723af518e247f134ff9a8c16
--- /dev/null
+++ b/test/tatoeba-test.ina-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dee405839f139ac78d9134eced27d44c4597cf795c46b055a1bbe1e33462946a
+size 73509
diff --git a/test/tatoeba-test.ina_Latn-lad_Latn.tsv b/test/tatoeba-test.ina_Latn-lad_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6ddc638e9085bdb29f045371c4ccca8ba6f79439
--- /dev/null
+++ b/test/tatoeba-test.ina_Latn-lad_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5d7886942ed888c12fbfc73426597dc27477df37bdb600f84d64bd8108bfe66
+size 16210
diff --git a/test/tatoeba-test.ina_Latn-lfn_Latn.tsv b/test/tatoeba-test.ina_Latn-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..69dab2761ca8644e094eeabbf1db6ce3689c522c
--- /dev/null
+++ b/test/tatoeba-test.ina_Latn-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:475f38d702fd3233d505e62571e01109e1f17c615ec849a59429d39c6f8edaa9
+size 48484
diff --git a/test/tatoeba-test.ina_Latn-tlh_Latn.tsv b/test/tatoeba-test.ina_Latn-tlh_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..80c65fa5e15193ae24f1667923ce6c008e887a1b
--- /dev/null
+++ b/test/tatoeba-test.ina_Latn-tlh_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a1913bb65333da6bc2e1fa3f88bef3c2d849496c9eb50746858d2e6e69fd722
+size 12311
diff --git a/test/tatoeba-test.ind-zsm_Latn.tsv b/test/tatoeba-test.ind-zsm_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6ac2e95a78d54228a25da8e212dcc7c0c57edaf4
--- /dev/null
+++ b/test/tatoeba-test.ind-zsm_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1f7149313a63a5990887359995a6c69fa0ccdc08cb88794a1498e25e8dab645
+size 23472
diff --git a/test/tatoeba-test.isl-ita.tsv b/test/tatoeba-test.isl-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..305d32e05808ff3e64c651f8326f8031ea1a2a0e
--- /dev/null
+++ b/test/tatoeba-test.isl-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:604b95b6262e6e31d4f191e76b2b18f6ad2714df8fdf9face525c4521c7ed140
+size 16140
diff --git a/test/tatoeba-test.isl-jpn.tsv b/test/tatoeba-test.isl-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..34cfbf3d420debce101779fc7cc67ca6c8c1ee6a
--- /dev/null
+++ b/test/tatoeba-test.isl-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9073181697080de58e13d82ab72de0b6d663d4bdade0767ffadf2055c0a5766a
+size 21926
diff --git a/test/tatoeba-test.isl-jpn_Hira.tsv b/test/tatoeba-test.isl-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..166370d43663242c905562915b875830ea02574c
--- /dev/null
+++ b/test/tatoeba-test.isl-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e3d3078c063022bc866c9a34df7cd2a098c6acb98cb99ab4e4a344eb8e734e1
+size 18260
diff --git a/test/tatoeba-test.isl-spa.tsv b/test/tatoeba-test.isl-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7d3be3021c182e69815135dfba32f033016a5118
--- /dev/null
+++ b/test/tatoeba-test.isl-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ada2541871f4db36965f44a4ce4a5e4d1e13330cf81ff4801d133d3a552fd55
+size 13806
diff --git a/test/tatoeba-test.ita-cmn_Hans.tsv b/test/tatoeba-test.ita-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..71f581b50b6253dff6b7e372d3ca1324cbf632b2
--- /dev/null
+++ b/test/tatoeba-test.ita-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b509977cbdd8a84bf60331ac707f1001a93c8d514db4668a1440c91474e7cad1
+size 138936
diff --git a/test/tatoeba-test.ita-cmn_Hant.tsv b/test/tatoeba-test.ita-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8501495ecbbd17079f81baf88f0f1fdc41f514e1
--- /dev/null
+++ b/test/tatoeba-test.ita-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab407fe50ab06181bd36d505dd2888f769154b5930baa2771742257c6401349f
+size 72270
diff --git a/test/tatoeba-test.ita-ind.tsv b/test/tatoeba-test.ita-ind.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..61bd38391465c33395dde1da9fbcbfc8a8e3db03
--- /dev/null
+++ b/test/tatoeba-test.ita-ind.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d81af8f71e00423a61dc1a0eb20fbef498f10a12e60e7f15167dabedab832170
+size 27058
diff --git a/test/tatoeba-test.ita-ita.tsv b/test/tatoeba-test.ita-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9166d0b87f3370a014e6ce984090a8db502bf3e6
--- /dev/null
+++ b/test/tatoeba-test.ita-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6fb8b0d6351b56eecd66896cad2ee4f00170b5c2153a8526111ac55d66e8f622
+size 79961
diff --git a/test/tatoeba-test.ita-jpn.tsv b/test/tatoeba-test.ita-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f8ff91d2b361442cfbbb309e659bd3e71175944b
--- /dev/null
+++ b/test/tatoeba-test.ita-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82ae67757fe3aabe923233ebdec16449cf76489c943a34a66dea3d954d6fd497
+size 273616
diff --git a/test/tatoeba-test.ita-jpn_Hani.tsv b/test/tatoeba-test.ita-jpn_Hani.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..51c5ebbbf3c15af118c18d0148e7aac70160d7d1
--- /dev/null
+++ b/test/tatoeba-test.ita-jpn_Hani.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:662cec10f3a406daf9d2697eca05fc1855428f019f19928f9ddf541108680400
+size 27233
diff --git a/test/tatoeba-test.ita-jpn_Hira.tsv b/test/tatoeba-test.ita-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2ed4c434ae0763c7cd4d185043672bb1da29cf74
--- /dev/null
+++ b/test/tatoeba-test.ita-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d39f127095dfc79fa156deb8e240813ed792fc34c28ec4b32bc204a85ee0767f
+size 241534
diff --git a/test/tatoeba-test.ita-lat.tsv b/test/tatoeba-test.ita-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..60d1bc0c5fa2ede020d4a48c9b8ffd5e92ca2997
--- /dev/null
+++ b/test/tatoeba-test.ita-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db448e9acae9c6615bbab4f89d9b49be2632f30504a16efdba544889ee0c89ca
+size 132924
diff --git a/test/tatoeba-test.ita-lit.tsv b/test/tatoeba-test.ita-lit.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e44d73fd6c24c4db9a94021ee9efe0887134e249
--- /dev/null
+++ b/test/tatoeba-test.ita-lit.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15a33b73b78574b7542ed1c4d322a885cc0d739b3238f423b6d78ba4687d45c2
+size 16502
diff --git a/test/tatoeba-test.ita-msa.tsv b/test/tatoeba-test.ita-msa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2f2095477b2767386d63acd4dc2c9d53a3d55c2d
--- /dev/null
+++ b/test/tatoeba-test.ita-msa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:017a6f2c7ca136afc29db68c94d8aeb0881643ae525705dca09f544fffdbd0c9
+size 32292
diff --git a/test/tatoeba-test.ita-nds.tsv b/test/tatoeba-test.ita-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c1fa4012bd71197fd027d6c55f398220113215f3
--- /dev/null
+++ b/test/tatoeba-test.ita-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6564ea6803dc4a2355f80259ef8628cbcfb5f5554067dc0ee5324ca945c752f3
+size 21384
diff --git a/test/tatoeba-test.ita-nld.tsv b/test/tatoeba-test.ita-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..763aea45b318cbbd9fa6eb62ce66c7e3f03b2082
--- /dev/null
+++ b/test/tatoeba-test.ita-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe3cf66e8cd000cd09da9a105db95735eba05c6e120bd6350ef18757ef7dac28
+size 173198
diff --git a/test/tatoeba-test.ita-nor.tsv b/test/tatoeba-test.ita-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..001eb4c1ee8897abf98b0ec3023f4e7b053b7357
--- /dev/null
+++ b/test/tatoeba-test.ita-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6406b8363b3451fe132a2d9427ccef1a847393cabc85512acc8097539f709ef5
+size 102588
diff --git a/test/tatoeba-test.ita-pms.tsv b/test/tatoeba-test.ita-pms.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b1ca24342913168de94b9f09dbcad55d658ada5e
--- /dev/null
+++ b/test/tatoeba-test.ita-pms.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55bd3456e4b45516ae70a205bf564ea98c86f2b36fc9678fa81d6b7f0ef0f165
+size 18864
diff --git a/test/tatoeba-test.ita-pol.tsv b/test/tatoeba-test.ita-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..77f1848bfc5f0c5a081bffd58bd6b04ca042c528
--- /dev/null
+++ b/test/tatoeba-test.ita-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9a4fd78df3bb84accab37422701e97c00f9e865bed19a7e047cf46d63bdbf46
+size 94013
diff --git a/test/tatoeba-test.ita-por.tsv b/test/tatoeba-test.ita-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c39da5ae9061caa9cc571aee18f97824708d7638
--- /dev/null
+++ b/test/tatoeba-test.ita-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69ebd0704374852102b82321e150e9ce0af61a9b188bafa2a8be74b572891526
+size 274134
diff --git a/test/tatoeba-test.ita-ron.tsv b/test/tatoeba-test.ita-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3c0ca8c7f03013b4d57c20d686c96d78b3a91260
--- /dev/null
+++ b/test/tatoeba-test.ita-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09a365731509a0c2fcc655ccb690ea43a490ccc62fa9ef501084a8b72f07e8f
+size 68917
diff --git a/test/tatoeba-test.ita-rus.tsv b/test/tatoeba-test.ita-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8a0fca0ac2af0c70622467e0ce57aa1f722e197e
--- /dev/null
+++ b/test/tatoeba-test.ita-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9fd219a41cc8d848e18b1c3982a980b969753ae4fd5e1a68c593bcc5e4e7a9f
+size 994450
diff --git a/test/tatoeba-test.ita-spa.tsv b/test/tatoeba-test.ita-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3a17d3a54a4c910e63b2f2a587c8aea5d255bd76
--- /dev/null
+++ b/test/tatoeba-test.ita-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8556a91c97644eeaa26c20526a3ec32600f9aa98b79dab1e89d7a5d48e54a0a
+size 381788
diff --git a/test/tatoeba-test.ita-swe.tsv b/test/tatoeba-test.ita-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cbe7569e02f372f78f392463e8b9c11902f3ba49
--- /dev/null
+++ b/test/tatoeba-test.ita-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c83723a8b1bc1b27c6703e91aeecc35cf931b484b9ec306ac3b96d6b673ecc6f
+size 51351
diff --git a/test/tatoeba-test.ita-toki.tsv b/test/tatoeba-test.ita-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..faf320ae64be9c6b66539ddf03170664216a1c36
--- /dev/null
+++ b/test/tatoeba-test.ita-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9816814e7d9c95b7de78d81ae7de49a1955efff71f46e6e1ebcec2b8ce9f1f5
+size 17073
diff --git a/test/tatoeba-test.ita-tur.tsv b/test/tatoeba-test.ita-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..706824b6ff3ca0b83c85f3a889475512c147d105
--- /dev/null
+++ b/test/tatoeba-test.ita-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d27bc6e3337887021a7b17e66596da3cee17e8ceafea04d99b17098d103d8820
+size 899367
diff --git a/test/tatoeba-test.ita-ukr.tsv b/test/tatoeba-test.ita-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..093ed06edd384ff19d8a6951332b0024e80bfa10
--- /dev/null
+++ b/test/tatoeba-test.ita-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92c864b8fb8269e0c86ed3ec1816bac15e7faf4c6ae3e7b529af724c55c38153
+size 373870
diff --git a/test/tatoeba-test.ita-vie.tsv b/test/tatoeba-test.ita-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1c406a9f32716facbedce82a88d92e88086c2f3e
--- /dev/null
+++ b/test/tatoeba-test.ita-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38d103539bc1953e90f7c188e3cd36f5e48ca34593bfe7b41a90c69c28098676
+size 21599
diff --git a/test/tatoeba-test.ita-yid.tsv b/test/tatoeba-test.ita-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..01c68eaa988b893665504f77f88fcec82a6940c4
--- /dev/null
+++ b/test/tatoeba-test.ita-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:200dab1b72ff32f9697cdec6941c680f59119504002b110edba3e7bd0fe86985
+size 16857
diff --git a/test/tatoeba-test.ita-zho.tsv b/test/tatoeba-test.ita-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a2b566e6cc1609ce96eaec72b8ebaa61ec6fe9da
--- /dev/null
+++ b/test/tatoeba-test.ita-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20c25206829b60cf271714b8c23f4881bbe4ef1ef058ee1578ce7c1f223d704d
+size 217098
diff --git a/test/tatoeba-test.jbo-jpn.tsv b/test/tatoeba-test.jbo-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bea96b13e489edc198697354055296ea4c7fd103
--- /dev/null
+++ b/test/tatoeba-test.jbo-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2747aa1c4db4d50b80fc828dc50873e2419ca0c72820bf63c7954fa9a96c3789
+size 81231
diff --git a/test/tatoeba-test.jbo-rus.tsv b/test/tatoeba-test.jbo-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5772e2391f9d593abe11e4fcb0bf38f39c7c6d7c
--- /dev/null
+++ b/test/tatoeba-test.jbo-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:62223ec6f729296363ccaa3a0fe214b89c3adbbffb2dd4af1afaa382add35cfe
+size 109668
diff --git a/test/tatoeba-test.jbo-spa.tsv b/test/tatoeba-test.jbo-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5707b7bf99262b8671ab85da1435f1182a7ebd0c
--- /dev/null
+++ b/test/tatoeba-test.jbo-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38746e3ba0b5f41d35d3297c99c72ec27b21b7e2792837c8f248e8b029123d5c
+size 107509
diff --git a/test/tatoeba-test.jbo-swe.tsv b/test/tatoeba-test.jbo-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ba1cc292f52887b64ca192942bf9e6382d210529
--- /dev/null
+++ b/test/tatoeba-test.jbo-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0eb0b407ac396cf433e82f0e2a43279755f9cd065a4934618c61440f409e2172
+size 15034
diff --git a/test/tatoeba-test.jbo-zho.tsv b/test/tatoeba-test.jbo-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8073d2fba64d6f7c99b1b06784d7a90f08bc385e
--- /dev/null
+++ b/test/tatoeba-test.jbo-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abef2ce023ad3c7b4c7df5c6678ecaeb55fddb9ab2f71c3407d5e7c86cde023d
+size 36339
diff --git a/test/tatoeba-test.jbo_Latn-cmn_Hans.tsv b/test/tatoeba-test.jbo_Latn-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..245f747673b631b135a059c519f9fff7534440e5
--- /dev/null
+++ b/test/tatoeba-test.jbo_Latn-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:876128d3891a01872f2c9cbeaa9b59c21097024dabaa2649de55325081f62991
+size 20111
diff --git a/test/tatoeba-test.jbo_Latn-cmn_Hant.tsv b/test/tatoeba-test.jbo_Latn-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..66ffd4b2d81f41dcbc0badb7e7bd2b0b68edf148
--- /dev/null
+++ b/test/tatoeba-test.jbo_Latn-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a6c0aec476725c430706d8c70574340a2ff58a73dd1c557ca0fb5e4e1caf678
+size 15670
diff --git a/test/tatoeba-test.jbo_Latn-jpn_Hira.tsv b/test/tatoeba-test.jbo_Latn-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..89ee2dc1616b6a4bf5ce996731d067b5376cb598
--- /dev/null
+++ b/test/tatoeba-test.jbo_Latn-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6084475e13fc80391170db4636f3baf625d0981f86aa7b28bc6314f509f2bc8
+size 72892
diff --git a/test/tatoeba-test.jpn-jpn.tsv b/test/tatoeba-test.jpn-jpn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bcc66d689c29d0895ad8d904a10ee6abf6f23b0e
--- /dev/null
+++ b/test/tatoeba-test.jpn-jpn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5585725d189fcee251d64b0bf38c14e28c0c65fb089e8d9f0c59caa75a836c9
+size 74694
diff --git a/test/tatoeba-test.jpn-kor.tsv b/test/tatoeba-test.jpn-kor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4bb13a914a85298d441f6b642b8a7a78adb208b7
--- /dev/null
+++ b/test/tatoeba-test.jpn-kor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32facf7cefd38b959148c42564820b8ed2afccb263e02a6193e049b0c7481ab6
+size 64304
diff --git a/test/tatoeba-test.jpn-lit.tsv b/test/tatoeba-test.jpn-lit.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..718e0dae6f0e68632b6f0939a2bd2148af661140
--- /dev/null
+++ b/test/tatoeba-test.jpn-lit.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4922368615f68a00a3c386944b89ae249b44ef6e824b979fdb6febee982b74b3
+size 21568
diff --git a/test/tatoeba-test.jpn-mar.tsv b/test/tatoeba-test.jpn-mar.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..75129b1989e132f5da98127b610a984287e9b4c2
--- /dev/null
+++ b/test/tatoeba-test.jpn-mar.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87e813189b566182b5b76233696919b51ffafb252bffeb14cb0314a570b8e94a
+size 43170
diff --git a/test/tatoeba-test.jpn-msa.tsv b/test/tatoeba-test.jpn-msa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b9043983bfe61f49276ff2a58950f43b531d6ebd
--- /dev/null
+++ b/test/tatoeba-test.jpn-msa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3b031d8bc1be07e733b01b93f5aeb05cc4e76990954a0943d83d282271fc005
+size 242969
diff --git a/test/tatoeba-test.jpn-nds.tsv b/test/tatoeba-test.jpn-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..730a1c9529848836b57d050702bc9413337c5d3f
--- /dev/null
+++ b/test/tatoeba-test.jpn-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48cfba1944d13e4be5d62738636b138d52c97ede63ad9d3db707a72c088e0168
+size 34566
diff --git a/test/tatoeba-test.jpn-nld.tsv b/test/tatoeba-test.jpn-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e1721b47d815e774e5445ded4d352eeb34e5d7e8
--- /dev/null
+++ b/test/tatoeba-test.jpn-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e0c2e460cd1773e2516ec24ea1c6828119c7eea4f22accfd56ed9995ec3d4e3
+size 311942
diff --git a/test/tatoeba-test.jpn-nor.tsv b/test/tatoeba-test.jpn-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0248fc9c254f62f4aadfb361b51ee056b0035675
--- /dev/null
+++ b/test/tatoeba-test.jpn-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b53861440c35c3ce0a91a48fda3b1e499072c376e46eaa185846cc2cf72b1b10
+size 104634
diff --git a/test/tatoeba-test.jpn-pol.tsv b/test/tatoeba-test.jpn-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..512f189349f442de7319c5888a7abf095eb1a140
--- /dev/null
+++ b/test/tatoeba-test.jpn-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b67ba51c4268075d70c32622332f623a47831c3e80670c1762321db4ba39ec76
+size 1044875
diff --git a/test/tatoeba-test.jpn-por.tsv b/test/tatoeba-test.jpn-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e22e9ea4458cc8ef766acbccd8c60b9e3b521bc9
--- /dev/null
+++ b/test/tatoeba-test.jpn-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f1ca170d3615e3937db6785a7c545559b8845b7e518bba2f1dec72a8f665a47
+size 193304
diff --git a/test/tatoeba-test.jpn-rus.tsv b/test/tatoeba-test.jpn-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bcaac420fe5e92551f4135b482a8cd79e9c9cd00
--- /dev/null
+++ b/test/tatoeba-test.jpn-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f52ebde508e778444834d462b4f3630312458b86cc70bff822fffe19713b153
+size 1259417
diff --git a/test/tatoeba-test.jpn-spa.tsv b/test/tatoeba-test.jpn-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..73d25d1144442d01b0fb5a47a2b57d05047a5e4e
--- /dev/null
+++ b/test/tatoeba-test.jpn-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b816c354cd704a0b9287088ff67d28a9943b45ab30f74644eb68197ba082955e
+size 1065342
diff --git a/test/tatoeba-test.jpn-swe.tsv b/test/tatoeba-test.jpn-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5f247d4b38a5b761f1795271b8e659bd150d7c80
--- /dev/null
+++ b/test/tatoeba-test.jpn-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5627260a30276525e622322af34aafcea1c2281dc55106f7b7a470cd9aff8dee
+size 82973
diff --git a/test/tatoeba-test.jpn-tlh.tsv b/test/tatoeba-test.jpn-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..64fdf5c9f17894f98ccc7004a7aac72d15ecbee2
--- /dev/null
+++ b/test/tatoeba-test.jpn-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e2ba4ea1ddaae3acbe9ee26f477fbb9eb3ea693df9d3e81a72be489abeef775
+size 65852
diff --git a/test/tatoeba-test.jpn-toki.tsv b/test/tatoeba-test.jpn-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..207ace6a284b14cc48b59fa4b2c2b2a3059a59f1
--- /dev/null
+++ b/test/tatoeba-test.jpn-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58441da1c5b3a3bb885f86e778e5914b0fdc3d87c9f1b3eb603ed578939b3dc1
+size 20860
diff --git a/test/tatoeba-test.jpn-tur.tsv b/test/tatoeba-test.jpn-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cccc0acd0363bc47ea10a165b689da5548aa200b
--- /dev/null
+++ b/test/tatoeba-test.jpn-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52437c282ff7c84a1ba00d3e219f998d799993b3e4f1df2b944726faba95916d
+size 78766
diff --git a/test/tatoeba-test.jpn-ukr.tsv b/test/tatoeba-test.jpn-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..72ed2de3f652418478cc5d392ea627c147f9ba45
--- /dev/null
+++ b/test/tatoeba-test.jpn-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4cd5a64ba14a7fa947a1c58cff2f6c7673c7f28c9fd7bad0a5fbf740ca15f573
+size 48744
diff --git a/test/tatoeba-test.jpn-vie.tsv b/test/tatoeba-test.jpn-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7435f288b9946945c77f0c6d23cb0edb8e36ef40
--- /dev/null
+++ b/test/tatoeba-test.jpn-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9eb980ce7e4e58a55d2fe1c1144b3f7af4fc70b88f43d7f05c0caf45b679baf
+size 122810
diff --git a/test/tatoeba-test.jpn-zho.tsv b/test/tatoeba-test.jpn-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bc2182294662a300461c04236239276d17f20071
--- /dev/null
+++ b/test/tatoeba-test.jpn-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b42f8a3624ba485618ec8f7e6c80366ae340fbe4d44dd6f51f8292c01e69f0e3
+size 244984
diff --git a/test/tatoeba-test.jpn_Hani-cmn_Hans.tsv b/test/tatoeba-test.jpn_Hani-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..67f0a0f12229a17f9b09286ecc0601293e9fed0e
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hani-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f052c4ea03a4559c407b88009e708190f1e0ffd610a3b395eca53aaf95a7a89
+size 18856
diff --git a/test/tatoeba-test.jpn_Hani-nld.tsv b/test/tatoeba-test.jpn_Hani-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bc3b6b48c1bdadde4c41c35f31c5a559b367e031
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hani-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1fc76db381707afcc8421b15b4d008e75dc9afb677ad9e5e317d8f7104a411d9
+size 35303
diff --git a/test/tatoeba-test.jpn_Hani-pol.tsv b/test/tatoeba-test.jpn_Hani-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0d0db743d078db29bdb8fe976a3e02bc9cd73f39
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hani-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25bea620b0bf0b934ad22a03db96328b12f3db13636462315de6588507f84014
+size 104212
diff --git a/test/tatoeba-test.jpn_Hani-por.tsv b/test/tatoeba-test.jpn_Hani-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..46a6fccef5ee173c9c34888f70c9b3f24829a935
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hani-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d2ade5d82450521d745c842b30cb05a6c17495afbb5260dbc80c47ad51b45c8
+size 30661
diff --git a/test/tatoeba-test.jpn_Hani-rus.tsv b/test/tatoeba-test.jpn_Hani-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1dbaf33ae3865639dd1ff56b769ea7e68efcc463
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hani-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:182fb4bc2821437af3d7123573d3d65ed5c351e8d5df44f2ba3acc30396d451b
+size 100811
diff --git a/test/tatoeba-test.jpn_Hani-spa.tsv b/test/tatoeba-test.jpn_Hani-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dfec6222e3a90730f1f2fbf6ae67b2084104de91
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hani-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:853adcfdf3162664597d3a0c970a99c149ebdb142fab6f418f7e830a1d4ea5d5
+size 102042
diff --git a/test/tatoeba-test.jpn_Hira-cmn_Hans.tsv b/test/tatoeba-test.jpn_Hira-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..313f5163cac729e49bb4d1fdcb24c13035485883
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fe0fb65f73720460b1083241f6b1d23e828b78aa60e7394e79c4bac14d282c5
+size 125687
diff --git a/test/tatoeba-test.jpn_Hira-cmn_Hant.tsv b/test/tatoeba-test.jpn_Hira-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c4ea867d73385c738d3b1323751b152df1a2a3e7
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d280822344fe39ffd98294dcee0aba2e6b4c45e6cde93309b1e0b0ed2061d314
+size 65875
diff --git a/test/tatoeba-test.jpn_Hira-ind.tsv b/test/tatoeba-test.jpn_Hira-ind.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e5a30b6c00af22345a1814f050fc342ec9bd1963
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-ind.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b96287cfc3c49d024088ba6ac4d6b561e3993bbc5fb76a5a2a148bb206c52a6
+size 209281
diff --git a/test/tatoeba-test.jpn_Hira-jpn_Hira.tsv b/test/tatoeba-test.jpn_Hira-jpn_Hira.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..346909c740848c81fe7e472ae7d6d61e196291db
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-jpn_Hira.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:842b3fce273bd90eee12e885355b4801ba94c767f2cbc5b7f62c6140ef3b8719
+size 60247
diff --git a/test/tatoeba-test.jpn_Hira-kor_Hang.tsv b/test/tatoeba-test.jpn_Hira-kor_Hang.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d5cb8fa8cab6e57eee36130806bffb34dbf498b9
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-kor_Hang.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5656bd9ef021f448a8df7f7978b8fc0c19acd65a53557ea558d050ba2f17832b
+size 53616
diff --git a/test/tatoeba-test.jpn_Hira-lit.tsv b/test/tatoeba-test.jpn_Hira-lit.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..395e5a229c95c37bae51277920b9f557e9597e32
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-lit.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8e1aa28ca63408df9aebf9144e02892bb4d32d1ac26a9bb18d9bb709f2396b1
+size 18266
diff --git a/test/tatoeba-test.jpn_Hira-mar.tsv b/test/tatoeba-test.jpn_Hira-mar.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..75c041cb19d4937582371eac5025f8957558b27a
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-mar.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b77c11bdbe53488bf80ec785f68e2fa129d3e880f26dd6dedc843a5228dedb2
+size 31703
diff --git a/test/tatoeba-test.jpn_Hira-nds.tsv b/test/tatoeba-test.jpn_Hira-nds.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9e612e3909ba51a8ad7cba0e601d4d32c31cb134
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-nds.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdca53864c1f3050d86a883d0a4194a4616b1b430349e7387a304791e1ee7f5f
+size 29378
diff --git a/test/tatoeba-test.jpn_Hira-nld.tsv b/test/tatoeba-test.jpn_Hira-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..192716963a74493ef1dd6ee43443dadda727f841
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de6621d1a651a841e7df00a0e87f9b2b030268b77f50b8bb6ea5ad655643027c
+size 269634
diff --git a/test/tatoeba-test.jpn_Hira-nob.tsv b/test/tatoeba-test.jpn_Hira-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1e57ddd90c3185689a46862bd635bf43732899ab
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ac4cbbada8b0263802a987e38be50c31277d4c0a1791ad751facab705a78d6b
+size 83793
diff --git a/test/tatoeba-test.jpn_Hira-pol.tsv b/test/tatoeba-test.jpn_Hira-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4a20ba128f18083b49375e2207f806a744a6f2e4
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51f26432e087f45e9e8d73a6ebde020c6696f313a90ed2830b49abf745b25558
+size 924931
diff --git a/test/tatoeba-test.jpn_Hira-por.tsv b/test/tatoeba-test.jpn_Hira-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..679aba377b22f931a613d5cbce9cf5cd5b91b836
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:050fb3ad85c70d3c8b7fcb0a4579e8e2718ad7502860486fbc63d178a65aabeb
+size 157262
diff --git a/test/tatoeba-test.jpn_Hira-rus.tsv b/test/tatoeba-test.jpn_Hira-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..871199c1643da446fa3cb794f28406300bddb316
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23a7793eb36abc83173f4af2e21315e2f2a775a0806b333aae3660d3a6508152
+size 1119177
diff --git a/test/tatoeba-test.jpn_Hira-spa.tsv b/test/tatoeba-test.jpn_Hira-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f5197a6ffe484a4b99de7858f3e7919ebbce2bca
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:973acf681f56fefb133714e29e7e8e2f7c9dc9287ba7195bd9c1d76fe5bb9125
+size 938050
diff --git a/test/tatoeba-test.jpn_Hira-swe.tsv b/test/tatoeba-test.jpn_Hira-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3cdbdf44abebc2048f67381ad620860a1cadee0b
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b015bf06d7c07f9999503f264daf0301aea624ff9b5b32d35ea730bc488a65e
+size 70646
diff --git a/test/tatoeba-test.jpn_Hira-tlh_Latn.tsv b/test/tatoeba-test.jpn_Hira-tlh_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..71e8f889be8e4443d7331af672ab85f47836d23e
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-tlh_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b99e73968bf0cb12caae243e2ba8da6e0c0783cc5192432d9536bea95958a472
+size 62553
diff --git a/test/tatoeba-test.jpn_Hira-tur.tsv b/test/tatoeba-test.jpn_Hira-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..72ee52e9087379d787f1b29e59a99fe9d34ca0a1
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98543ebdf3e7f5944d44cb29809639d3d8bbdd27c82ba5a14c70e897444f59db
+size 71606
diff --git a/test/tatoeba-test.jpn_Hira-ukr.tsv b/test/tatoeba-test.jpn_Hira-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0ef92e67cf9f69bc28f46e9ec4ad5d109e50f308
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1385a050bbaeb0b96ce86c06a742ca0c6ece2979f9aa65c511632b1fc58d2b75
+size 41595
diff --git a/test/tatoeba-test.jpn_Hira-vie.tsv b/test/tatoeba-test.jpn_Hira-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..09f391cc023eeb5ac9ebb7ad941c5a8a9cc088f9
--- /dev/null
+++ b/test/tatoeba-test.jpn_Hira-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1be62866745e616300b6de9e9098ff3190b39e1ae6bc802805560b4ac2174ada
+size 107208
diff --git a/test/tatoeba-test.jpn_Kana-rus.tsv b/test/tatoeba-test.jpn_Kana-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c07e3871ec6a00c654686994ca04011f977d01a5
--- /dev/null
+++ b/test/tatoeba-test.jpn_Kana-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a57031dd9ee83b3f23ab538746942515417cd2d866c56e131dccc1213a206454
+size 37932
diff --git a/test/tatoeba-test.jpn_Kana-spa.tsv b/test/tatoeba-test.jpn_Kana-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a9dd32de9b53b8e334c2e5be54e494276c1e741c
--- /dev/null
+++ b/test/tatoeba-test.jpn_Kana-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0f97053dbe18a4c5effb646046e99217260908e5bc9133a8795cb56cff209c8
+size 24577
diff --git a/test/tatoeba-test.kab-kab.tsv b/test/tatoeba-test.kab-kab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4b3d7922f9cd063a993c94741eb9027266fdccac
--- /dev/null
+++ b/test/tatoeba-test.kab-kab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:071f91ad4b6bf875a7f812f6114be8e144c4477239ae71d0b640fa20848948d7
+size 92832
diff --git a/test/tatoeba-test.kab-rus.tsv b/test/tatoeba-test.kab-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e8551fbf3cbfdff667a3fa08ca5f209b90b51b09
--- /dev/null
+++ b/test/tatoeba-test.kab-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3feb4b4407ffc5d22353caa05135132def64c26f2c75d5546e39414e8d7a3485
+size 39530
diff --git a/test/tatoeba-test.kab-spa.tsv b/test/tatoeba-test.kab-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..40ac957080e442c502a88c45372e7b8ce302c9d6
--- /dev/null
+++ b/test/tatoeba-test.kab-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a73f0532590d20213aa7f97ac257cb324d85de9a979c93ef79ce473a8caab16e
+size 70630
diff --git a/test/tatoeba-test.kat-rus.tsv b/test/tatoeba-test.kat-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..105de0dead53344cc2b31495a96a21a8fe8db4b4
--- /dev/null
+++ b/test/tatoeba-test.kat-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89c7da3a1815bcdf832823633b031c300076a090740c62c950a5b868fbb11a54
+size 91485
diff --git a/test/tatoeba-test.kaz-rus.tsv b/test/tatoeba-test.kaz-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8e4fa2a2c9858b0bbd75f5ea34128e7c72084309
--- /dev/null
+++ b/test/tatoeba-test.kaz-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29f5930204dc73fc46d949914b0a87e667bae56b7bd3a75ca7cae8a07c720e64
+size 337480
diff --git a/test/tatoeba-test.kaz_Cyrl-rus.tsv b/test/tatoeba-test.kaz_Cyrl-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d9b06dfc94c241232b1dbb2b0c5542fd14a77752
--- /dev/null
+++ b/test/tatoeba-test.kaz_Cyrl-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93bb65df4386087f7a801f28f3d46b0548db695a366c2a249686a7dfd6337b28
+size 337316
diff --git a/test/tatoeba-test.khm-spa.tsv b/test/tatoeba-test.khm-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ea836716f7c28f43270ea21cfdf6176907d124a9
--- /dev/null
+++ b/test/tatoeba-test.khm-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c30bf54dfe9b6089abb906583a72093e7def80f62509eda5f078f12611ba097
+size 147017
diff --git a/test/tatoeba-test.kor-rus.tsv b/test/tatoeba-test.kor-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8b6f0f2a6b9d7ee2a07720269b51c2a9ef7e8274
--- /dev/null
+++ b/test/tatoeba-test.kor-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92fa5e02359558e7ccd25b793d095bb202ab9e4d31ef66f96e42215e044ab144
+size 23083
diff --git a/test/tatoeba-test.kor-spa.tsv b/test/tatoeba-test.kor-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ee9780ae0d91e37ec694f5b570793db78bd824a1
--- /dev/null
+++ b/test/tatoeba-test.kor-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14dd352d514fcbf00799bab9ecfdbce605817181f72f605c48aa500b6fa1dde0
+size 83074
diff --git a/test/tatoeba-test.kor-zho.tsv b/test/tatoeba-test.kor-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..671b800b3079a94258bc4107e77e4cef44f1fe2a
--- /dev/null
+++ b/test/tatoeba-test.kor-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4327a0bd3e2bfe4277b5667844a9602c228b26b910d6c274388e24e0f60420c9
+size 37105
diff --git a/test/tatoeba-test.kor_Hang-cmn_Hans.tsv b/test/tatoeba-test.kor_Hang-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6c887f9cb16fa8b5249664c87f2aab204ac73410
--- /dev/null
+++ b/test/tatoeba-test.kor_Hang-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8ee9e02cd65232ce185f4240cb11d91cad32e990a45a7a3e791d4b0539f510e
+size 18974
diff --git a/test/tatoeba-test.kor_Hang-rus.tsv b/test/tatoeba-test.kor_Hang-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8e1ce85effb38b9858ac042980bd2143eb0f014b
--- /dev/null
+++ b/test/tatoeba-test.kor_Hang-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f092f496c49359b27be680f5959bcfb11558f28822e82bdf00d9f8df75a6b65
+size 23000
diff --git a/test/tatoeba-test.kor_Hang-spa.tsv b/test/tatoeba-test.kor_Hang-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bb6168c3e45ef5b1443ccfbbaf91a60ded14636a
--- /dev/null
+++ b/test/tatoeba-test.kor_Hang-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7980d02ac2904d046f5f374a6719534e240c93b2807821b096057a62edbe7383
+size 81877
diff --git a/test/tatoeba-test.kzj-msa.tsv b/test/tatoeba-test.kzj-msa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e8eb5df9305036dd32081a03bf71e3de4363b47d
--- /dev/null
+++ b/test/tatoeba-test.kzj-msa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d6528956f6d030b118888d2bbab30d1770d2d4387e7fa71fb731baa28602879
+size 45722
diff --git a/test/tatoeba-test.kzj_Latn-zsm_Latn.tsv b/test/tatoeba-test.kzj_Latn-zsm_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a2b70d6f085040b47c5ee5c929bd463d1e0e18cb
--- /dev/null
+++ b/test/tatoeba-test.kzj_Latn-zsm_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a919a7afe168747fc1f96aa6a161ba19c83dabbadb51c981f20d919ea83184c7
+size 43012
diff --git a/test/tatoeba-test.lad-lat.tsv b/test/tatoeba-test.lad-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..604e5db54f5277fc2ca8d5a765d3b8ea0d918195
--- /dev/null
+++ b/test/tatoeba-test.lad-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:669cee079f62aa327ac52a2d1059a7e0bde1901d6d74f2b7eb65d13f5bd5efdf
+size 14669
diff --git a/test/tatoeba-test.lad-lfn.tsv b/test/tatoeba-test.lad-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0382383577bbec85b7262fe4af3ab007ee085fb9
--- /dev/null
+++ b/test/tatoeba-test.lad-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f43baeb2eb407eeb3ae7f0e90a2bf9649a81bac60b8536243212587f7561e92
+size 23973
diff --git a/test/tatoeba-test.lad-spa.tsv b/test/tatoeba-test.lad-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4865a3b95102d81250ed36ba70dd7a7d39ba6be6
--- /dev/null
+++ b/test/tatoeba-test.lad-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98c50b2c666bc56e9d785a2b6a58d7d8a709086c3b5fc074d920015362cbf1b4
+size 17333
diff --git a/test/tatoeba-test.lad-yid.tsv b/test/tatoeba-test.lad-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0c550e8c4a72691ef00436a5fbc5134fd8c8012a
--- /dev/null
+++ b/test/tatoeba-test.lad-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:885125269f0e03cd7d946e704fec55a5021da7b80db386a2eaeb237e3697180b
+size 55133
diff --git a/test/tatoeba-test.lad_Latn-lfn_Latn.tsv b/test/tatoeba-test.lad_Latn-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6f21341909c813499ed2917999362361f49fcd34
--- /dev/null
+++ b/test/tatoeba-test.lad_Latn-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:168458d7ced2609b2cfe32eec2b61d21879c5cd801583aa75db8e069d710e3eb
+size 13412
diff --git a/test/tatoeba-test.lad_Latn-spa.tsv b/test/tatoeba-test.lad_Latn-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..19255fcc02a1104043ba08c3b35c76c047bcc1a6
--- /dev/null
+++ b/test/tatoeba-test.lad_Latn-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:747948f4859df9760db9c13fc655376398e0683d782d0424ff72b3823c6a1338
+size 14299
diff --git a/test/tatoeba-test.lad_Latn-yid.tsv b/test/tatoeba-test.lad_Latn-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..94898480790e9f00ba442c5f5c3e3f1ea2aaef87
--- /dev/null
+++ b/test/tatoeba-test.lad_Latn-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcf5ed65ad6af9ef7cea21dd375c9471b26ce7ae9d3b8335f7e7bc63e1011755
+size 37993
diff --git a/test/tatoeba-test.lat-lat.tsv b/test/tatoeba-test.lat-lat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7d91d6e3b911be6fcc2fc4546a54aafa05d6c24d
--- /dev/null
+++ b/test/tatoeba-test.lat-lat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0f06ba4f5336f27ff300c08f79139bee867bcca40571ce20e3d0efb921c0073
+size 20141
diff --git a/test/tatoeba-test.lat-lfn.tsv b/test/tatoeba-test.lat-lfn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..56af9e5c84294ad6a70225d8fa905024e69b489c
--- /dev/null
+++ b/test/tatoeba-test.lat-lfn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8137b1ca42f11eaa98bfd74fdbf54d4e276881712461d4e13a9dda3b0b78d8cc
+size 22981
diff --git a/test/tatoeba-test.lat-nld.tsv b/test/tatoeba-test.lat-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ecb3444fede80ccdad67b503598a030b7282bbcc
--- /dev/null
+++ b/test/tatoeba-test.lat-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34a8066633c0e1c9db7eb765c3024d0fc1fe18eaa756713b44592ebe7afa403e
+size 24902
diff --git a/test/tatoeba-test.lat-nor.tsv b/test/tatoeba-test.lat-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..38793d55811c759a1096b9bc480042f8f4d406de
--- /dev/null
+++ b/test/tatoeba-test.lat-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3366f0de2080b0d720bd91cd026efcd937af6dff0c3e56e18b887567841cdca4
+size 23606
diff --git a/test/tatoeba-test.lat-pol.tsv b/test/tatoeba-test.lat-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d25597db1b8eaae5524aab8f6bc5fe51df1a43fd
--- /dev/null
+++ b/test/tatoeba-test.lat-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c25afbd3320f085f57d065711a17433e15387f6ccbc835ad0213917a2f2d0366
+size 61028
diff --git a/test/tatoeba-test.lat-por.tsv b/test/tatoeba-test.lat-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7d0f04a4b6d8dfb513376636b8f252eeaecb7fbc
--- /dev/null
+++ b/test/tatoeba-test.lat-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2f10f26b6f3a257c9ce2377ab1830913b7d5cc378ebcdd75c32555d33beea13
+size 626036
diff --git a/test/tatoeba-test.lat-rus.tsv b/test/tatoeba-test.lat-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..98f0e940676762170266aa13fc5a8d5936eef5a0
--- /dev/null
+++ b/test/tatoeba-test.lat-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dbbe63074c072281fc414e5f3716566549f1350120f7f5c02389aafaa7f176fe
+size 95593
diff --git a/test/tatoeba-test.lat-spa.tsv b/test/tatoeba-test.lat-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c2e70abc506fe9ef61fa688f279debe5adea9f7f
--- /dev/null
+++ b/test/tatoeba-test.lat-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2767d81957d0890cbd717f708c0009c6bf76997966c19061e98f4eccc3dfbd9b
+size 346603
diff --git a/test/tatoeba-test.lat-tlh.tsv b/test/tatoeba-test.lat-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1c8626cc1d710568887c63c30d345bf5eb32a12d
--- /dev/null
+++ b/test/tatoeba-test.lat-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9cbd7f567911c4cb3ae3d82997c9cba0a1667dac5882b775ee4f254925f0bcc4
+size 14111
diff --git a/test/tatoeba-test.lat-ukr.tsv b/test/tatoeba-test.lat-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0a24bf2ea414f777904d36a087d8f4bd916db71f
--- /dev/null
+++ b/test/tatoeba-test.lat-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3e7264826d6a40de2afe9ae765d9698e9096717d75a09506af32f1b511424e4
+size 29975
diff --git a/test/tatoeba-test.lat-yid.tsv b/test/tatoeba-test.lat-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..06078fe8c8c1f57c464dec8ed58154458e5f22d7
--- /dev/null
+++ b/test/tatoeba-test.lat-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e0a074f16cb3195fe8232d3b590cbf8d4904ad59f588204c16752f093aaf3c4
+size 41312
diff --git a/test/tatoeba-test.lat_Latn-lfn_Latn.tsv b/test/tatoeba-test.lat_Latn-lfn_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..daef2e1edb58ab4805a9a633e47a603247c65669
--- /dev/null
+++ b/test/tatoeba-test.lat_Latn-lfn_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb73e8a8881b479f7e6d0c547e485b6755c459b045eb4f68fcfb823705abdab0
+size 19102
diff --git a/test/tatoeba-test.lat_Latn-por.tsv b/test/tatoeba-test.lat_Latn-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0331f3ab2fc983661bc339d0d4164f8ce9e3f859
--- /dev/null
+++ b/test/tatoeba-test.lat_Latn-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84c244fe968c4f50081786eef06a36135b58f926bb70b46b6aa6009435692c30
+size 625922
diff --git a/test/tatoeba-test.lav-rus.tsv b/test/tatoeba-test.lav-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..13046d9276bdea9ffd1cd22cc9c14f12186e68c5
--- /dev/null
+++ b/test/tatoeba-test.lav-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24002469646180f59b68ce5b68e93d03013096b680d6b1f18f568a4ee7f852c8
+size 22259
diff --git a/test/tatoeba-test.lfn-por.tsv b/test/tatoeba-test.lfn-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..941a10053dfd7a823338cd829945aa88a6957db1
--- /dev/null
+++ b/test/tatoeba-test.lfn-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9eb05c67dee9e99e05af5079076197def8be992658c3ef2a37c7831c9c1d19b4
+size 280715
diff --git a/test/tatoeba-test.lfn-rus.tsv b/test/tatoeba-test.lfn-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1a0c21be2d30b2bd26e9207a23d31aaf6160d1f0
--- /dev/null
+++ b/test/tatoeba-test.lfn-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ffdb9186ad0b1b1911af1461f18a2a68ca0028ef4e0e133d515785d82631f45a
+size 25529
diff --git a/test/tatoeba-test.lfn-spa.tsv b/test/tatoeba-test.lfn-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0d91e058c159884c165076cea39297aeecc895f1
--- /dev/null
+++ b/test/tatoeba-test.lfn-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e83771927a795d46711f89ba1d8cdf3567456b94284553824a652199f4aee7df
+size 21163
diff --git a/test/tatoeba-test.lfn-yid.tsv b/test/tatoeba-test.lfn-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..12c19a9210f8a7240a478b571f1c9334d84012ef
--- /dev/null
+++ b/test/tatoeba-test.lfn-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:842eb93a716cd982a98cef413b77bfbff87ac829be8540ff17f1abf1bcdf23da
+size 93617
diff --git a/test/tatoeba-test.lfn_Cyrl-por.tsv b/test/tatoeba-test.lfn_Cyrl-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a7c97128e7bdf6fd219fb1845b414eea5acfdd42
--- /dev/null
+++ b/test/tatoeba-test.lfn_Cyrl-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e785d2e8c1cd11c96f7051077ccb81346489d419bb6b70afd19ee346f0a99821
+size 28489
diff --git a/test/tatoeba-test.lfn_Latn-por.tsv b/test/tatoeba-test.lfn_Latn-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..41fc5000a4454731d7996df559a05835da442890
--- /dev/null
+++ b/test/tatoeba-test.lfn_Latn-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0153e10fb0fdeec3f3f4a3fa59aea51bdea561c5b0d122951f5dd378b0b21a3f
+size 252226
diff --git a/test/tatoeba-test.lfn_Latn-yid.tsv b/test/tatoeba-test.lfn_Latn-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b50d9457a78cc45a561f4550e855239e29f8d8cb
--- /dev/null
+++ b/test/tatoeba-test.lfn_Latn-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6714d10a48296cdea1eafbf2b25b03e129c552c047b756336ca8e384e929112
+size 81847
diff --git a/test/tatoeba-test.lit-pol.tsv b/test/tatoeba-test.lit-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0fd65846d6bb5c312de1a4baf603bbc541125493
--- /dev/null
+++ b/test/tatoeba-test.lit-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f734c754c462eb7a9d14bb1fb7d65fcb3de05aaaef2b609e0a5003de58a1652
+size 126233
diff --git a/test/tatoeba-test.lit-rus.tsv b/test/tatoeba-test.lit-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b1195935328800ffd1ce03511dd8028b57986409
--- /dev/null
+++ b/test/tatoeba-test.lit-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc198508687aeadae9a8edd70ac8ba5ff2e8f32b2d7d8b18970a87bf69d80dbf
+size 328689
diff --git a/test/tatoeba-test.lit-spa.tsv b/test/tatoeba-test.lit-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b334c1e0b87c8e3196d89f6de0d9e907d481fa90
--- /dev/null
+++ b/test/tatoeba-test.lit-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94d66148806f9cdfd7e67aedea3729aa89ffaf3758c4e9d948adbc30b7502106
+size 29135
diff --git a/test/tatoeba-test.lit-tur.tsv b/test/tatoeba-test.lit-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fd1c0bfdd0e025b0d19ffd1e55808a0c0c84a3dd
--- /dev/null
+++ b/test/tatoeba-test.lit-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fb7c95c958363c741b735937cb052f62e1b9f4c7c4a43b635983614d63ccc74
+size 102387
diff --git a/test/tatoeba-test.ltz-nld.tsv b/test/tatoeba-test.ltz-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..267c96bbae2d998f7aa979dec799102747313afd
--- /dev/null
+++ b/test/tatoeba-test.ltz-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39f5363d6605cbba6539dd4624224f8ef05f9c31f19d5e837fcc555707b44640
+size 17279
diff --git a/test/tatoeba-test.mkd-spa.tsv b/test/tatoeba-test.mkd-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5cb055fc0a4a8a93281400b4a6cb0631607cc9fa
--- /dev/null
+++ b/test/tatoeba-test.mkd-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8909500a2edd906b4765509a15c9ce3d785c0cb03c830efd23ed88c16d3f8df1
+size 14994
diff --git a/test/tatoeba-test.msa-msa.tsv b/test/tatoeba-test.msa-msa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a28799e27110f90dee4bfe02708704b4ffba9e08
--- /dev/null
+++ b/test/tatoeba-test.msa-msa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbeffad866e82214e7b0ed8dec06fcd96940fa2d998eb1cce6124ecf4db22a08
+size 81386
diff --git a/test/tatoeba-test.msa-spa.tsv b/test/tatoeba-test.msa-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b6144337712f79bf1251b2c489e2ba61e26bb671
--- /dev/null
+++ b/test/tatoeba-test.msa-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b6e0a2febc2bab8fd9a62fd4d4ad270fadb8d548709c72f58253336b4fe16bc
+size 16593
diff --git a/test/tatoeba-test.msa-zho.tsv b/test/tatoeba-test.msa-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a218eaa8244f0f733abf14913723d1aa31c01198
--- /dev/null
+++ b/test/tatoeba-test.msa-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7110179f5336c0b30db4cbd5f89a3f289f2e7289e66a660337f9af8c42f360f
+size 33792
diff --git a/test/tatoeba-test.nds-nld.tsv b/test/tatoeba-test.nds-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2ed78d6a183a2a3fd121753ce1f840956b4c7bb8
--- /dev/null
+++ b/test/tatoeba-test.nds-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23ffcab3760ac46e60b7322be38784d7286315c85a94a637dae3c0db996a3bb8
+size 117605
diff --git a/test/tatoeba-test.nds-por.tsv b/test/tatoeba-test.nds-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..51eeccf1ab7f8b545ca152ab469951eb3f44f8b6
--- /dev/null
+++ b/test/tatoeba-test.nds-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbb82f2f44a7cf9b886de13356682a234acf5304aacae63f6e512809dba3245a
+size 12995
diff --git a/test/tatoeba-test.nds-rus.tsv b/test/tatoeba-test.nds-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e86e62eeb9facf5efc0b1d16d5640cb84dafa9df
--- /dev/null
+++ b/test/tatoeba-test.nds-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f95601c493a697b02be2264d5bed9e58d4f67602b391372730f4d3fdcb502439
+size 78622
diff --git a/test/tatoeba-test.nds-spa.tsv b/test/tatoeba-test.nds-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b09e6deee318803e931c86b4ff6da96474d1c53e
--- /dev/null
+++ b/test/tatoeba-test.nds-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e253b420e820d9ea2167faa835ae1bbaa7347bc87556cb56bb1cc70629317dd
+size 59877
diff --git a/test/tatoeba-test.nld-cmn_Hant.tsv b/test/tatoeba-test.nld-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..91a47fd10cf71347c651452ec6c9c28537bad861
--- /dev/null
+++ b/test/tatoeba-test.nld-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8eec73670877d1e9cfa16fd1f4ad22cca79e8fb759b7b6fcd87647cf3ca2d08e
+size 130492
diff --git a/test/tatoeba-test.nld-nld.tsv b/test/tatoeba-test.nld-nld.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..62ea00488428632b1c6f5a008db5a8050dc0cd08
--- /dev/null
+++ b/test/tatoeba-test.nld-nld.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:facfc04bdb44ad2e039644b0a3bf124949284ad9d74600e15f7d0bd2a7c531bc
+size 74782
diff --git a/test/tatoeba-test.nld-nor.tsv b/test/tatoeba-test.nld-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..881a228f5e4a60b2a166040a00f16e0dcb4020f7
--- /dev/null
+++ b/test/tatoeba-test.nld-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c62c5fe81353de37bffbf3d65cba4de8f8da73f3e25a0f0d3e04a1678fb7f0db
+size 15232
diff --git a/test/tatoeba-test.nld-pol.tsv b/test/tatoeba-test.nld-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a6d845c67c33d2e1e9fe73f8cb4d6275f9fe72d3
--- /dev/null
+++ b/test/tatoeba-test.nld-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c8d7b2441c9320d267ac51d4140cc7b9a30331a53b8d08159a11844e4f97463
+size 87695
diff --git a/test/tatoeba-test.nld-por.tsv b/test/tatoeba-test.nld-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f311354a607fdfc4b21c6c0096b06fba2da50164
--- /dev/null
+++ b/test/tatoeba-test.nld-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd527d2f7ee668dc496f9c47eee9ac121c56392e5a41dc381ad73786ca61451e
+size 185289
diff --git a/test/tatoeba-test.nld-ron.tsv b/test/tatoeba-test.nld-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d43d36088852a047dd0c6ba22563976737224e6e
--- /dev/null
+++ b/test/tatoeba-test.nld-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59bc2bc7adb4d076eda0d2ffcbe58b03ffdf0ef8d9306e93ec5132f290575e35
+size 171326
diff --git a/test/tatoeba-test.nld-rus.tsv b/test/tatoeba-test.nld-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fe51f151457f8800af01fc981c6d35ccf72500d8
--- /dev/null
+++ b/test/tatoeba-test.nld-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b31159dfd51d51e054cac89e6c9a296b79728fbe06c6039c847bc09c3148c6b
+size 242444
diff --git a/test/tatoeba-test.nld-spa.tsv b/test/tatoeba-test.nld-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..df3991170ab6fcc45dde2b210b8f83ce24dddf1d
--- /dev/null
+++ b/test/tatoeba-test.nld-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d34fbeea3c036a5023f06492102a2bc7d99f49c47380b4d29db8840487aca0f
+size 816229
diff --git a/test/tatoeba-test.nld-toki.tsv b/test/tatoeba-test.nld-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9b4167ed3e0d8f5e36d8a2ee1ab15d948c356d7e
--- /dev/null
+++ b/test/tatoeba-test.nld-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0d16b9362886738cf001473ee5aa844b2d1a2fe39d4527761b2937f04dfa24f
+size 43655
diff --git a/test/tatoeba-test.nld-tur.tsv b/test/tatoeba-test.nld-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f743d8f73415768079ba88ee4c369996a2fd0dd6
--- /dev/null
+++ b/test/tatoeba-test.nld-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:571297a5f3c888bc674280c0746db221b892fd7a11bdb9b15b334f5183a87f80
+size 184277
diff --git a/test/tatoeba-test.nld-ukr.tsv b/test/tatoeba-test.nld-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1f02bcf88d116ee56683554d913851a17daa32eb
--- /dev/null
+++ b/test/tatoeba-test.nld-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35653559717f35e21b7d0c0f5a295a9ed1c3d428702f7a03bab51ae446700d8f
+size 775813
diff --git a/test/tatoeba-test.nld-zho.tsv b/test/tatoeba-test.nld-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cd616d10437129c0ceeb7c0b5001c3c36377f652
--- /dev/null
+++ b/test/tatoeba-test.nld-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa84411ad6dbf490095ff78367b47cdda7706a6def19e32c410abaeefb71705f
+size 140786
diff --git a/test/tatoeba-test.nno-nob.tsv b/test/tatoeba-test.nno-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..db46fe94089ad29afbad55d63829b9f51e80cc59
--- /dev/null
+++ b/test/tatoeba-test.nno-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f77c20941b9c4a5421d1f83ca59a801edc4a9d430a3653799aac1329b3f8d5a6
+size 31927
diff --git a/test/tatoeba-test.nob-nno.tsv b/test/tatoeba-test.nob-nno.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8175346020b6eca046796e283b61e7a7d56443af
--- /dev/null
+++ b/test/tatoeba-test.nob-nno.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b21b86338bb25379b6116ca51a80b0881e5472e1acf528a64141ff093fdf8d9
+size 31873
diff --git a/test/tatoeba-test.nob-rus.tsv b/test/tatoeba-test.nob-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..35922158674065e16a5e977a09259d311ae74af8
--- /dev/null
+++ b/test/tatoeba-test.nob-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80ba6d6d45c0a7f09b55cb47366e8d56aa0c2cda4e4e508f52c81ef41234d60f
+size 170914
diff --git a/test/tatoeba-test.nob-spa.tsv b/test/tatoeba-test.nob-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..550831a3c3eccb58b1782e50e572b1529d8c5195
--- /dev/null
+++ b/test/tatoeba-test.nob-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccd54f4039152ea8df58bb5a7e3b68b6351278162c92786e033f2d20fe2758af
+size 74321
diff --git a/test/tatoeba-test.nob-swe.tsv b/test/tatoeba-test.nob-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..41eaaa1f3afa177136309faf4495f104e9b26581
--- /dev/null
+++ b/test/tatoeba-test.nob-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d5269bb5e655fb02e7b05e6b96f00d92e70dd115611c4fd211056490c1005f8
+size 38582
diff --git a/test/tatoeba-test.nor-nor.tsv b/test/tatoeba-test.nor-nor.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..59aea2ff4191abd8ecf56b2cae9a5ee0efcc53fa
--- /dev/null
+++ b/test/tatoeba-test.nor-nor.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d310cbe243b6100c00f482c8c78f96b99f7f81362c694962c5d60b4fcac4308
+size 67160
diff --git a/test/tatoeba-test.nor-pol.tsv b/test/tatoeba-test.nor-pol.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..341deb3787805795122d860fe365fd1889be20b3
--- /dev/null
+++ b/test/tatoeba-test.nor-pol.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae679e0ceae45237606f257c7b064660c92016425e07b134cc81e71bb6df4c38
+size 21984
diff --git a/test/tatoeba-test.nor-por.tsv b/test/tatoeba-test.nor-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4f7934c95ccf312e97dd12f709e690e1314896df
--- /dev/null
+++ b/test/tatoeba-test.nor-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a9bb8ba230520bf377944d187b259f36b8e5d6f72601312ac2bffa8dbf8b894
+size 43546
diff --git a/test/tatoeba-test.nor-rus.tsv b/test/tatoeba-test.nor-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b85e842db217693a32416389a9c9704a9b12fe52
--- /dev/null
+++ b/test/tatoeba-test.nor-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0edcde99bb11bd295866c8e4fc947ff4c04fca0273a153cb051b3ab26bfa71e
+size 171086
diff --git a/test/tatoeba-test.nor-spa.tsv b/test/tatoeba-test.nor-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8ec2b49dcd9b3c5dcf9137b786f0527a6fbac944
--- /dev/null
+++ b/test/tatoeba-test.nor-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ce5c746aabe0708c17af66b08733894572b0930a165bcc2c5e11f244002feb1
+size 79106
diff --git a/test/tatoeba-test.nor-swe.tsv b/test/tatoeba-test.nor-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..07f20650e3704fbbf2045c62b04e4753e8d6fe5a
--- /dev/null
+++ b/test/tatoeba-test.nor-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:885ec65572342404bf47b1f17ec8e8b4c75538e1afd3f5a4f310daaf02c71858
+size 39044
diff --git a/test/tatoeba-test.nor-ukr.tsv b/test/tatoeba-test.nor-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..26737d9b86d072687c424ffaa001dc1e7dbf3299
--- /dev/null
+++ b/test/tatoeba-test.nor-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da52ea12277b97531971ad24b0cef1aeaf04bc9efe126b459bf697bb80676717
+size 59241
diff --git a/test/tatoeba-test.nor-zho.tsv b/test/tatoeba-test.nor-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..96af45c8dbad7b5434471fb68ceb42d275dc824f
--- /dev/null
+++ b/test/tatoeba-test.nor-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10719c18640de1b83fae81b7eef5f0642e92cd03abf0d4667a8d7bf9db5c1d3c
+size 16243
diff --git a/test/tatoeba-test.orv-ukr.tsv b/test/tatoeba-test.orv-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ee5a5f2197592617edc4640993265b8ec9d0b72d
--- /dev/null
+++ b/test/tatoeba-test.orv-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca51ad26539effd83d16ed191e6402af58dc3b8280afc5ae4ca369f2336e2646
+size 95946
diff --git a/test/tatoeba-test.ota-tur.tsv b/test/tatoeba-test.ota-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b1fc5b0c7eec2b7ca8f59a985ba8b0da8ef96d0e
--- /dev/null
+++ b/test/tatoeba-test.ota-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9307d718dc0efa89852bc19d17d3b411c5db217f1bf1731742150ce8da43b5a7
+size 29366
diff --git a/test/tatoeba-test.pol-cmn_Hans.tsv b/test/tatoeba-test.pol-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ef55811290a9118bc2a4d1b3b5fa46807841796c
--- /dev/null
+++ b/test/tatoeba-test.pol-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc06f135322ccf6fb5641822a834ca7f2e6d69a22a7c35be39520e39f7f1bbca
+size 30094
diff --git a/test/tatoeba-test.pol-cmn_Hant.tsv b/test/tatoeba-test.pol-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4d417aedb6517316861ef966cada3eeedb9db7bc
--- /dev/null
+++ b/test/tatoeba-test.pol-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41aba34e1f91bcafe5e25b2b9324d1336ad91d2835eb3a0e07b52082ed3713f2
+size 50414
diff --git a/test/tatoeba-test.pol-por.tsv b/test/tatoeba-test.pol-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..917ade5e00bd3202301e77d5bcde1fc401e32393
--- /dev/null
+++ b/test/tatoeba-test.pol-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54bed9a7a916e673b05517c9f4f709cf544643e70035ac86e4584b5e2bea14ef
+size 54377
diff --git a/test/tatoeba-test.pol-rus.tsv b/test/tatoeba-test.pol-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0c7cd0fa361c54dd45c6c1de95c12c0f948b0746
--- /dev/null
+++ b/test/tatoeba-test.pol-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe4a0683dbb5178b43d834c3bf3abce3c719c4b68a337f734d42f1fd15f24482
+size 333617
diff --git a/test/tatoeba-test.pol-spa.tsv b/test/tatoeba-test.pol-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..aec367c2e17b9d8ebcf5654b14917f2af30f190e
--- /dev/null
+++ b/test/tatoeba-test.pol-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7d9c56b98c22c0c2a47643872d95c3d8773910120021bae4a869e87069f78a9
+size 195296
diff --git a/test/tatoeba-test.pol-swe.tsv b/test/tatoeba-test.pol-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0e6fd0e2f4aff72afe1d6c83ee748411ebab4ad8
--- /dev/null
+++ b/test/tatoeba-test.pol-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3cf5eae25f48584cd6ec40a936e8791108cbc5fad157f91c57e34e6b21adbad7
+size 100023
diff --git a/test/tatoeba-test.pol-tur.tsv b/test/tatoeba-test.pol-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c9087d0de71aed90ebee1bfa17acc4a989ba6da0
--- /dev/null
+++ b/test/tatoeba-test.pol-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93000b8b52f09e3a9d45201187f5dc0fc3d73fcc685502ec63837c6e1ae92422
+size 66267
diff --git a/test/tatoeba-test.pol-ukr.tsv b/test/tatoeba-test.pol-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ed43f9e11b7dea0b20a816540040e5e4d4788983
--- /dev/null
+++ b/test/tatoeba-test.pol-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a95be65d4d7a99521339cdecf5d786570001944d212f5ef2a2ba9ed71f8fd617
+size 199615
diff --git a/test/tatoeba-test.pol-zho.tsv b/test/tatoeba-test.pol-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..99636889b4b3f8844dbe4ac3160babb13ddac451
--- /dev/null
+++ b/test/tatoeba-test.pol-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e7e9369572c75e318d7067012ef118aeac84652106af8b8f2f12566162a152a
+size 82974
diff --git a/test/tatoeba-test.por-cmn_Hans.tsv b/test/tatoeba-test.por-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b8c1976afa130d746a0443d8293703f6564a488f
--- /dev/null
+++ b/test/tatoeba-test.por-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:967108f3afb601314d8bc8ab647b863401362b75ae68a7aba67e60750e2af4b3
+size 48909
diff --git a/test/tatoeba-test.por-cmn_Hant.tsv b/test/tatoeba-test.por-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2d76813c3dc15177ca50d5ec8b0077d03dfd3610
--- /dev/null
+++ b/test/tatoeba-test.por-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a6ae0ec64560bcf9001f46c3f83893e26f93c9e3d66b30bf83ad5b5cad72a95
+size 26065
diff --git a/test/tatoeba-test.por-por.tsv b/test/tatoeba-test.por-por.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..72921fdfc4f7d06d7011d9307754df1f3242177c
--- /dev/null
+++ b/test/tatoeba-test.por-por.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c840236a1a3e5960275b834540e68cb8d3c4b451a7f0eb1b9dd31f585c75eeb
+size 204599
diff --git a/test/tatoeba-test.por-ron.tsv b/test/tatoeba-test.por-ron.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1b0e3283ded92fba45eb2cfbe4e7f19b32b7c269
--- /dev/null
+++ b/test/tatoeba-test.por-ron.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daefe3eb71e5724d8625f464899399c3cf629f4c71e7f915af1b62b78a5cebcc
+size 49155
diff --git a/test/tatoeba-test.por-rus.tsv b/test/tatoeba-test.por-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6519b54add57040d9090974f5c8e944c4d82e90d
--- /dev/null
+++ b/test/tatoeba-test.por-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fde482eacf15c7a532a6a07fe62ecc756b77a98535d843ba1a5e6affa02a9141
+size 1012258
diff --git a/test/tatoeba-test.por-spa.tsv b/test/tatoeba-test.por-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9fd32c26667f0b70d24a161c4cd7334951597815
--- /dev/null
+++ b/test/tatoeba-test.por-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58150df326c2d3affd7a18a9b35b7be83fe5728fd8eb4a2925f11304b4afdf6f
+size 944876
diff --git a/test/tatoeba-test.por-swe.tsv b/test/tatoeba-test.por-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fc57470d5c185309f76a802cd841d951a6067df7
--- /dev/null
+++ b/test/tatoeba-test.por-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f2fa630033bce36174c7e95f212c1c933a811f47cd42de7783a9186c0d435f4
+size 21549
diff --git a/test/tatoeba-test.por-tgl.tsv b/test/tatoeba-test.por-tgl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..6c4fecc2d901ca3e4f596d4ffaee5e9e4551950f
--- /dev/null
+++ b/test/tatoeba-test.por-tgl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f94b292b7e19f9095332865560dd4eaef6e60a8e7a47c45866c7afaf91d08211
+size 152019
diff --git a/test/tatoeba-test.por-toki.tsv b/test/tatoeba-test.por-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..92261544d2d2460ae19a004e8c2828ceba112e7b
--- /dev/null
+++ b/test/tatoeba-test.por-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:724a02e8df3560bf6e958b99c05d3e349c5e48cb22d998af00dcc2de118d7102
+size 159364
diff --git a/test/tatoeba-test.por-tur.tsv b/test/tatoeba-test.por-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7a1ab4a2b59a341c34a8e062fa69bf3bf72cb2ca
--- /dev/null
+++ b/test/tatoeba-test.por-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8dd01b51c40221334b41a3825e704a3f44f80f454c9ffd6ac869b9c5ccdb93a1
+size 125270
diff --git a/test/tatoeba-test.por-ukr.tsv b/test/tatoeba-test.por-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c51ddd25c4bd3d745015099ed322e6c8e5a31d80
--- /dev/null
+++ b/test/tatoeba-test.por-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67e293c0e41fd9ef14bfcc9371d317cfcb350e92190126d61e56b487afc49fa4
+size 285017
diff --git a/test/tatoeba-test.por-zho.tsv b/test/tatoeba-test.por-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a62fca85514c7576446bcea4675466970b5cc86f
--- /dev/null
+++ b/test/tatoeba-test.por-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9be539b6a06b48887f57ed5bc88ec42d72d00794730f10640de346daa388db3
+size 75621
diff --git a/test/tatoeba-test.ron-rus.tsv b/test/tatoeba-test.ron-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8596f206dd6bf70282111a874ba489c2ef062e0b
--- /dev/null
+++ b/test/tatoeba-test.ron-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f600a71104b5129c1299ebdc268936302b36b9edb2871beca323fec911bc73db
+size 66481
diff --git a/test/tatoeba-test.ron-spa.tsv b/test/tatoeba-test.ron-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..16a3c5916d93464ad621abac98b49a7d3a2dc3a8
--- /dev/null
+++ b/test/tatoeba-test.ron-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f195539b6f3c0f5fb8c5438b08cdfdd2dd13dd979c2e2405e223014ff1add6b
+size 137180
diff --git a/test/tatoeba-test.ron-tur.tsv b/test/tatoeba-test.ron-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c8040169ada7f5df3b1edc4dc91faaa88458a2b4
--- /dev/null
+++ b/test/tatoeba-test.ron-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fb7dd7038730a67f3847f21969ec1f236db0228d0437c45e371c5c48c1acf82
+size 203717
diff --git a/test/tatoeba-test.run-rus.tsv b/test/tatoeba-test.run-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..983b890c4a65c70ff0c6411a3c03fc7159f5b07d
--- /dev/null
+++ b/test/tatoeba-test.run-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b0597ce78624b3ff72284d99d170a21f35f1a7be63fda0340eeb82eac936c44
+size 86691
diff --git a/test/tatoeba-test.run-spa.tsv b/test/tatoeba-test.run-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a076d5599f8804ad9912570a10d0a66f9f28c694
--- /dev/null
+++ b/test/tatoeba-test.run-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04509917c1987677ad5dc457619f21648c083114f86d2f182ffae88051c101a9
+size 52277
diff --git a/test/tatoeba-test.rus-cmn_Hans.tsv b/test/tatoeba-test.rus-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..19168474416700b5fa084c6a45e9e4b24604c6fb
--- /dev/null
+++ b/test/tatoeba-test.rus-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c452de40ba86b4e515f51b2336fd850b08d7ff449b0230b20163b3891afc2e8c
+size 116284
diff --git a/test/tatoeba-test.rus-cmn_Hant.tsv b/test/tatoeba-test.rus-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..70d9f6e7f92d6644b7aeea848ad9d912cea271bb
--- /dev/null
+++ b/test/tatoeba-test.rus-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0209b9aa0770af0d30f0301a85d790038ca16cf5b6553bedbb33655ec1cf2a61
+size 74745
diff --git a/test/tatoeba-test.rus-rus.tsv b/test/tatoeba-test.rus-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ad38a1309a36eb6d6939f1311ba689770ae75251
--- /dev/null
+++ b/test/tatoeba-test.rus-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c7ba974c0ca14071623364e7e2c8d249737cdfcaadbbc693ea8f698ec30911a
+size 312967
diff --git a/test/tatoeba-test.rus-sah.tsv b/test/tatoeba-test.rus-sah.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9e9bad223e71e6fbc948f27725a10b345371a605
--- /dev/null
+++ b/test/tatoeba-test.rus-sah.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d8f2f6f8967b710541c0217af82c3f2d1929a49c607c1ba73ef0b5c66f94d50
+size 117992
diff --git a/test/tatoeba-test.rus-slv.tsv b/test/tatoeba-test.rus-slv.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..97908a77aa45e24ab633c60d421db090333809d0
--- /dev/null
+++ b/test/tatoeba-test.rus-slv.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bdf3d0bad556d012c0ccd80abc3df2b979ead7c2f343ab8fb5518ed5b1af171
+size 56930
diff --git a/test/tatoeba-test.rus-spa.tsv b/test/tatoeba-test.rus-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..2f21c392fc70892de440c08c5cdfdbc875e9a05a
--- /dev/null
+++ b/test/tatoeba-test.rus-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf9684ca69178f7d5b9a8f4520a7a48792c2b45b5feb22e765508841c7b7b16a
+size 1050372
diff --git a/test/tatoeba-test.rus-swe.tsv b/test/tatoeba-test.rus-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5031bf140ab6775879b51a52296a3c9fc44056b4
--- /dev/null
+++ b/test/tatoeba-test.rus-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:519f10bac33d6936b82ac1bf4fe34ea02896a0df2b09492fbf3a32cb077bd66a
+size 116623
diff --git a/test/tatoeba-test.rus-tat.tsv b/test/tatoeba-test.rus-tat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..897a90bb1682173ef2e63becab0ac438ee3301a6
--- /dev/null
+++ b/test/tatoeba-test.rus-tat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b00ab9565bcbca2e492faab7bb049295e489a33610f0a18cef2a275f48fb7619
+size 448786
diff --git a/test/tatoeba-test.rus-tlh.tsv b/test/tatoeba-test.rus-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..099d8ec2990846dd50271c935a65c564c9b1a5e9
--- /dev/null
+++ b/test/tatoeba-test.rus-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83c1f1935e64f200add4176cac8e0415c11605216b2d1b81151e668487cae3fd
+size 19143
diff --git a/test/tatoeba-test.rus-toki.tsv b/test/tatoeba-test.rus-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..899a4e429773a01dd032ed02a5e8b3cd66fc752f
--- /dev/null
+++ b/test/tatoeba-test.rus-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:342ce40c2e615acbb682a6a33bf0e064288a8fdc9234635ae945539b5a49c7e0
+size 86608
diff --git a/test/tatoeba-test.rus-toki_Latn.tsv b/test/tatoeba-test.rus-toki_Latn.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..810257c89a9522b54fc8f95af69f30a8e23c56b3
--- /dev/null
+++ b/test/tatoeba-test.rus-toki_Latn.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:976e3b9cadf3d9aa9f5735bd9f2797824da225dc425f2367cbe9682d7108e710
+size 86569
diff --git a/test/tatoeba-test.rus-tur.tsv b/test/tatoeba-test.rus-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1e799cae9ac7cd778546f429270e1c9529d9f66d
--- /dev/null
+++ b/test/tatoeba-test.rus-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77ca7a2908bb8fa1168578dbd5a6e0b3d897714d8c1d34af469162e27a9367d9
+size 517450
diff --git a/test/tatoeba-test.rus-uig.tsv b/test/tatoeba-test.rus-uig.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..adc699d1c7a61325ebff3abb71af4a6c3c12fa50
--- /dev/null
+++ b/test/tatoeba-test.rus-uig.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d40a382dfce29d9b8898bf5e94bf18afc898fbb8290a62cd0ac7965e04ce693
+size 66192
diff --git a/test/tatoeba-test.rus-uig_Arab.tsv b/test/tatoeba-test.rus-uig_Arab.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b2b12167cf44b5b4056f03f184e2ff5f2e049ec7
--- /dev/null
+++ b/test/tatoeba-test.rus-uig_Arab.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9287c51c324255fefe1546d23438c5c8280b64af4d1601bf58cbba91d9af52e5
+size 65448
diff --git a/test/tatoeba-test.rus-ukr.tsv b/test/tatoeba-test.rus-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..153af6c666979615d87b7d49729b35c7130af962
--- /dev/null
+++ b/test/tatoeba-test.rus-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8fe2e005f0cf04e24e096f59e4176bb66e9988b57cb59404a383bb279ffa561
+size 1107093
diff --git a/test/tatoeba-test.rus-vie.tsv b/test/tatoeba-test.rus-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..270f0ba3d0a2b5ba8ea9dae5f7d2fb421ea4da19
--- /dev/null
+++ b/test/tatoeba-test.rus-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9de04913f5d99ee8ca631f22239c338b144831d58aa648c3eff34bd9925cb1a2
+size 34414
diff --git a/test/tatoeba-test.rus-xal.tsv b/test/tatoeba-test.rus-xal.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..51b6240c3490dbd1b0d5e04e17b5a7d6e172ff8a
--- /dev/null
+++ b/test/tatoeba-test.rus-xal.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:685e435003f13fa6a45b109e6407201a7808fe433fb0ca1b599fa7d72cbd1cb1
+size 23410
diff --git a/test/tatoeba-test.rus-yue_Hans.tsv b/test/tatoeba-test.rus-yue_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..9214f3f4f4b8cee96d839f89becb1354bbaa29e7
--- /dev/null
+++ b/test/tatoeba-test.rus-yue_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0c0fdc57b3c415f0975648c975e9aff2d58ff9e0007b2421ac1d17755780b11
+size 23533
diff --git a/test/tatoeba-test.rus-zho.tsv b/test/tatoeba-test.rus-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..63fb4ced86193c11be637c4d5392c4a9f38d4810
--- /dev/null
+++ b/test/tatoeba-test.rus-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d5d4f2083d23bca93130e78aafd6a13b97e100c5df08070124093398d4ce3f0
+size 260121
diff --git a/test/tatoeba-test.slv-cmn_Hans.tsv b/test/tatoeba-test.slv-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c8a563fb896295c5c0b99820c7a6e577c5d27e31
--- /dev/null
+++ b/test/tatoeba-test.slv-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8a6a6e7950b052a9b2fae97b551df61a1c996dd1dcb282499750479f278582d
+size 54243
diff --git a/test/tatoeba-test.slv-ukr.tsv b/test/tatoeba-test.slv-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..45fce0b7dbe8eece3b9ee12433a8348112ded30a
--- /dev/null
+++ b/test/tatoeba-test.slv-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8de798c5682b639746e9ac826c962054da1332cc0904d9c595aa197cbbbacb5a
+size 60177
diff --git a/test/tatoeba-test.slv-zho.tsv b/test/tatoeba-test.slv-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..36dc2dd2f69a679a510ff342d642e0b66250243b
--- /dev/null
+++ b/test/tatoeba-test.slv-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f3e6fc74d5008206e256a16b2fef14594258f13da1e32bf7904d5a76a7a1327
+size 60602
diff --git a/test/tatoeba-test.spa-cmn_Hans.tsv b/test/tatoeba-test.spa-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a0d445997a2342447ff06ca0ba55806eb43a664a
--- /dev/null
+++ b/test/tatoeba-test.spa-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d69f9059c00ffb999193eb7e85f8d3247b8b1ca8b97f00120e36d3cc7933075c
+size 83817
diff --git a/test/tatoeba-test.spa-cmn_Hant.tsv b/test/tatoeba-test.spa-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4e78e3bd3835b6288d2fed6298d1e338b56c73f5
--- /dev/null
+++ b/test/tatoeba-test.spa-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d49d1d0b84ad29a9f3ff42bc3cc08481d619e133d153bfcc4d14e5280b955bd9
+size 105555
diff --git a/test/tatoeba-test.spa-spa.tsv b/test/tatoeba-test.spa-spa.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..dbee415c8854c5e85d0d1acd6cb2258be26f02ed
--- /dev/null
+++ b/test/tatoeba-test.spa-spa.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f8d8d72eb10d84943b088e6be57a27c24968de546ecef6aaf8e4b9d45765c6c
+size 231019
diff --git a/test/tatoeba-test.spa-swe.tsv b/test/tatoeba-test.spa-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0d2ddfc5e3f01440cee0b037bb2ebea750ea4a5d
--- /dev/null
+++ b/test/tatoeba-test.spa-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6c0a4cbceb9d69899d24813a41faba80e48f495e9e7c49cd87c6e6b4ec12540
+size 89887
diff --git a/test/tatoeba-test.spa-tat.tsv b/test/tatoeba-test.spa-tat.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0043b8350ea39de113ed4e72bdded77dd6b8ec3c
--- /dev/null
+++ b/test/tatoeba-test.spa-tat.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:024886f2d032f40f9a913cc5924a9b545dc2a85874b44bbabee65ee182bf40a2
+size 30965
diff --git a/test/tatoeba-test.spa-tgl.tsv b/test/tatoeba-test.spa-tgl.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f6546ef35582fcf7be490682219314aa33046321
--- /dev/null
+++ b/test/tatoeba-test.spa-tgl.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef217584b352fac3fea756ca89a20417cc65c4da5f17a27b760099d9fe5386d8
+size 51801
diff --git a/test/tatoeba-test.spa-tlh.tsv b/test/tatoeba-test.spa-tlh.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..a04496014b04d677fcb7607e01a0e4a54cdef82d
--- /dev/null
+++ b/test/tatoeba-test.spa-tlh.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18a8d6b1dccd4666a1c4fb51a5e5589b0f30acf9ff10965401a2b0ea80be418e
+size 20288
diff --git a/test/tatoeba-test.spa-toki.tsv b/test/tatoeba-test.spa-toki.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..fdf09d363de5c7945037873b9f7358d91732124c
--- /dev/null
+++ b/test/tatoeba-test.spa-toki.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e27d94fd6ba6465cdaa6234a9ae362bf0215b4b785e91bb224b56a55939379cc
+size 46015
diff --git a/test/tatoeba-test.spa-tur.tsv b/test/tatoeba-test.spa-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7d6846e3a7f830908ec1a4543df5c4a16d1baf0d
--- /dev/null
+++ b/test/tatoeba-test.spa-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:230ecda1414ecaa70e4c96834d6377c589e4985a33f7fa133db272bd548b714e
+size 743546
diff --git a/test/tatoeba-test.spa-ukr.tsv b/test/tatoeba-test.spa-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1bf3b57104fb147c33abe8d785623db93c9da64d
--- /dev/null
+++ b/test/tatoeba-test.spa-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aea907d46d14f1e432cb8269d86d424155fecec4d3b159dde0a75c9180951d73
+size 817523
diff --git a/test/tatoeba-test.spa-vie.tsv b/test/tatoeba-test.spa-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cab0b73dc2aadeffae197f847b7728ea7bec76ba
--- /dev/null
+++ b/test/tatoeba-test.spa-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53a33719fb6fda3fb4e75ae99395dd584d4dbc57551765476d3bfd60362126e0
+size 46565
diff --git a/test/tatoeba-test.spa-yid.tsv b/test/tatoeba-test.spa-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..5922b924caade6566ea10d1300eb6859c62eb0dd
--- /dev/null
+++ b/test/tatoeba-test.spa-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5aba646c93df2df65e533ffb47f535d17b628178a6784e01ce11385807830d57
+size 37495
diff --git a/test/tatoeba-test.spa-zho.tsv b/test/tatoeba-test.spa-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c9cb3c7fcf30f9a119a09df518161cc6879df197
--- /dev/null
+++ b/test/tatoeba-test.spa-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a1100edbb4fb4f03437c11cfbe5b106606db2197a71abddb446605eaa6c6980
+size 195859
diff --git a/test/tatoeba-test.srp_Cyrl-rus.tsv b/test/tatoeba-test.srp_Cyrl-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f08970ae28bf8f9e6f4f526a5bfca39af870e968
--- /dev/null
+++ b/test/tatoeba-test.srp_Cyrl-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e49e47230be7778a275a093047f9cacdbc761c021b1071a825f0b2293236034
+size 98682
diff --git a/test/tatoeba-test.srp_Cyrl-ukr.tsv b/test/tatoeba-test.srp_Cyrl-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d90ea3db82326359775d5fe22b1f60fe222ccbd0
--- /dev/null
+++ b/test/tatoeba-test.srp_Cyrl-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32fbe708df2c42180b0e6fe68af60c7e1c6426b68020a286e4b8307afde7017c
+size 19037
diff --git a/test/tatoeba-test.srp_Latn-ita.tsv b/test/tatoeba-test.srp_Latn-ita.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0b8bdc1259a76257ae583b406b1da27739d1dd3c
--- /dev/null
+++ b/test/tatoeba-test.srp_Latn-ita.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3a066781b2650197cc59025e657ef1f7740d5dcc534886d189e953b4241453a
+size 14324
diff --git a/test/tatoeba-test.srp_Latn-nob.tsv b/test/tatoeba-test.srp_Latn-nob.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..3fa757904309d8bb65d5e3c34321c906331ab319
--- /dev/null
+++ b/test/tatoeba-test.srp_Latn-nob.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:233df09cfc9566b88481d94ee24ebfdabc2d597ca12e1212d5560e0ed96284e5
+size 460044
diff --git a/test/tatoeba-test.srp_Latn-rus.tsv b/test/tatoeba-test.srp_Latn-rus.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e286af7bdd4368fdc891aacf62230384490bfc22
--- /dev/null
+++ b/test/tatoeba-test.srp_Latn-rus.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b2787bb816ba1c4c9ca3be85e2b7cec1f08d8e936d6483032b0f6824d58b558
+size 127298
diff --git a/test/tatoeba-test.srp_Latn-ukr.tsv b/test/tatoeba-test.srp_Latn-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c6facf1c08e9443fedbe40659c85b95b92e46e68
--- /dev/null
+++ b/test/tatoeba-test.srp_Latn-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:681356fe4ae66eb2092cc388f1fb775dc52e44162640fe83e69bc67a03cd005c
+size 25013
diff --git a/test/tatoeba-test.swe-cmn_Hans.tsv b/test/tatoeba-test.swe-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..26c86e93cb715c9490cab73894487738e9f54fa2
--- /dev/null
+++ b/test/tatoeba-test.swe-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa512090f0583d2a773ff8fc4d6fcbb6a27513263309eeafaf9c4c305899c7bb
+size 33354
diff --git a/test/tatoeba-test.swe-cmn_Hant.tsv b/test/tatoeba-test.swe-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ede62a64fdf34f7692c396961bf842ab15b866a8
--- /dev/null
+++ b/test/tatoeba-test.swe-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7156b305a0aa9fb50391d802ba31b665aae090725786d557226d4b42f7bf59d
+size 35044
diff --git a/test/tatoeba-test.swe-swe.tsv b/test/tatoeba-test.swe-swe.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..7f11247e37a335725b07c94c756c01b1a7c786a9
--- /dev/null
+++ b/test/tatoeba-test.swe-swe.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2dcc78d91584bc72fa101d7a28c2f5e7f1bef9052d1bc896c101ca3b106d94a
+size 72315
diff --git a/test/tatoeba-test.swe-tur.tsv b/test/tatoeba-test.swe-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cb4963cbc1501c6b0e66de949c52f9a118e9e8f9
--- /dev/null
+++ b/test/tatoeba-test.swe-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36cd06c628d0ee55c5e199f08e4d5d44a9449977baa1afbe7b3517218ec5f772
+size 14122
diff --git a/test/tatoeba-test.swe-zho.tsv b/test/tatoeba-test.swe-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..f2be492413fbfd2d11a18e8a19871ee4c4e5b5dd
--- /dev/null
+++ b/test/tatoeba-test.swe-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fee55f0d26da90b2991c9bed314fb560352ddc229f9470acb7eed73c3d0dc20a
+size 68783
diff --git a/test/tatoeba-test.tat-tur.tsv b/test/tatoeba-test.tat-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..de0968fa0ffab06b00bcc774cbd662ddf327beef
--- /dev/null
+++ b/test/tatoeba-test.tat-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf5c566a7828bd97aeb201ae4f2bb97436cd12ece6e39e1d37e4cf223e472770
+size 46968
diff --git a/test/tatoeba-test.tat-vie.tsv b/test/tatoeba-test.tat-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..924e1f0f9a009dc83f072fdd4d00cba2ef600ed4
--- /dev/null
+++ b/test/tatoeba-test.tat-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92d4b9804b31740a6ea5bd28cabe4defada3b7e08f3a1c9de6db241c5856059d
+size 30530
diff --git a/test/tatoeba-test.tlh-yid.tsv b/test/tatoeba-test.tlh-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..42095d37772724389fcd71cb5dc5c55b08c4d05f
--- /dev/null
+++ b/test/tatoeba-test.tlh-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c92f81a369a8cb3dc9577bfdfedc5618b44c1b157df14fe502346075aceaa5d
+size 23168
diff --git a/test/tatoeba-test.tlh-zho.tsv b/test/tatoeba-test.tlh-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e033b0e594510ce51fbd803c8bdab218b80e78a5
--- /dev/null
+++ b/test/tatoeba-test.tlh-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98a218eaa7128360c56061b5f899819c26a88e4bb835ef257a894e2cddc1b5ba
+size 31164
diff --git a/test/tatoeba-test.tlh_Latn-cmn_Hans.tsv b/test/tatoeba-test.tlh_Latn-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..65e282c465ba0c45113ec65d7ebde8218a0e0b96
--- /dev/null
+++ b/test/tatoeba-test.tlh_Latn-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7af6fc4c436fd488acfe47e8bf5b61ab64334308e75086f7b856df74389ef279
+size 17590
diff --git a/test/tatoeba-test.tlh_Latn-cmn_Hant.tsv b/test/tatoeba-test.tlh_Latn-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..8e49df1bd668d2de77be56c25ff7824054239390
--- /dev/null
+++ b/test/tatoeba-test.tlh_Latn-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb813728ecddde6aa24f1ae2db0986212e9f696c495477197c94fe5ab3e28801
+size 13574
diff --git a/test/tatoeba-test.tlh_Latn-yid.tsv b/test/tatoeba-test.tlh_Latn-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d56ed32c5c1a906029ae751fff32f0efc2735c12
--- /dev/null
+++ b/test/tatoeba-test.tlh_Latn-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfb02aefd5b46504662ceae46a3f51951c88802779eab6e85791f068983dc80d
+size 23118
diff --git a/test/tatoeba-test.tur-cmn_Hans.tsv b/test/tatoeba-test.tur-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..d3449384e5035b5ea22ed06fc3750a9511e6fb11
--- /dev/null
+++ b/test/tatoeba-test.tur-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f9e22bd5f5e1348c2567f2b2a94289ef071d956469d80aec6f58940f76856f7
+size 40855
diff --git a/test/tatoeba-test.tur-cmn_Hant.tsv b/test/tatoeba-test.tur-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..4754b5c166fb35a334c4c43b02f50108c57868bb
--- /dev/null
+++ b/test/tatoeba-test.tur-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d3f728ae75dd40892048818d3b15ee793a2cf0aaa3a815ef7a4ff34d8f9509f
+size 26193
diff --git a/test/tatoeba-test.tur-tur.tsv b/test/tatoeba-test.tur-tur.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..0c2f6c712d656d2517c5ddf9cfa14a77ff18f37e
--- /dev/null
+++ b/test/tatoeba-test.tur-tur.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8934e6517031305fc4118f0b2dcc94c410b7441be753780619f0cde840a65e6
+size 182115
diff --git a/test/tatoeba-test.tur-uig.tsv b/test/tatoeba-test.tur-uig.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..c963bf3155635d0c21e7c56576103eafa0390a11
--- /dev/null
+++ b/test/tatoeba-test.tur-uig.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bea99bab0149cb479d0bd4e005a01c5f29adddd48838f8d3a7fcffa34a4ff26
+size 150099
diff --git a/test/tatoeba-test.tur-ukr.tsv b/test/tatoeba-test.tur-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..11ce7ed7f2620ca0d59381c6e2b421123d6b08f0
--- /dev/null
+++ b/test/tatoeba-test.tur-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:caa4d4bc67c5e273b4ca9847a46fcc47a90a6c76fe357f68142f7e5b6c0b4746
+size 192375
diff --git a/test/tatoeba-test.tur-uzb.tsv b/test/tatoeba-test.tur-uzb.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..181004d03e7ff502f8abed342d51be5388fbf2b3
--- /dev/null
+++ b/test/tatoeba-test.tur-uzb.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4836b0789cb6839ea1bb529a202a24ec0bf0405813339f4a5afcc545fe809384
+size 14378
diff --git a/test/tatoeba-test.tur-zho.tsv b/test/tatoeba-test.tur-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..845a99288b2f83757e751bdd10a2891b00c3a36a
--- /dev/null
+++ b/test/tatoeba-test.tur-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fcdc85c61a8d50263c2fce6e69eeb5c5278fb5a1f2ef2942be1de9f484fb62e
+size 67341
diff --git a/test/tatoeba-test.uig-zho.tsv b/test/tatoeba-test.uig-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..81e4a2f891c75a96a80fab276db86612aece10c2
--- /dev/null
+++ b/test/tatoeba-test.uig-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2ff391356ccd060c46065d47cf5e8b4639cee4665f51dca8a0551a0a0f4ed47
+size 207018
diff --git a/test/tatoeba-test.uig_Arab-cmn_Hans.tsv b/test/tatoeba-test.uig_Arab-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..cef9791546ef7931c0869f48afb4f4cad4d6baa2
--- /dev/null
+++ b/test/tatoeba-test.uig_Arab-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9e12bc8aa2150542f9ba6f63fd8ed8bc2764ff713623a0a6e762c57d65511e9
+size 151405
diff --git a/test/tatoeba-test.uig_Arab-cmn_Hant.tsv b/test/tatoeba-test.uig_Arab-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..1e195ea74159bda97a267729bc1d3ba9da0e66eb
--- /dev/null
+++ b/test/tatoeba-test.uig_Arab-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:753c41aca5fafe6750451234198bef427dcbef7bb219aa7bac68c0b9607cbff9
+size 55221
diff --git a/test/tatoeba-test.ukr-cmn_Hans.tsv b/test/tatoeba-test.ukr-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..afa474355c603ad69e448112cd304dece161ec4e
--- /dev/null
+++ b/test/tatoeba-test.ukr-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f15739227e6b1ddb9b32ce07cda76adee1fcc11e078ad9d22105d8b46cd4a74
+size 79448
diff --git a/test/tatoeba-test.ukr-cmn_Hant.tsv b/test/tatoeba-test.ukr-cmn_Hant.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..bf79edee90099828821edff31d4ba369719fd242
--- /dev/null
+++ b/test/tatoeba-test.ukr-cmn_Hant.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4eec367c7919d88b5ad7aae56e95dfba7206d66e7c75736fcdc4b71679d636f
+size 42184
diff --git a/test/tatoeba-test.ukr-ukr.tsv b/test/tatoeba-test.ukr-ukr.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..b334fd64f2500b615528d70924ea7c44813dbd36
--- /dev/null
+++ b/test/tatoeba-test.ukr-ukr.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ea0547587ef8b5dc14063cf87e5e30b010fcc57d2bdc36cac18123ad3664fca
+size 73415
diff --git a/test/tatoeba-test.ukr-zho.tsv b/test/tatoeba-test.ukr-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..311756a5e3619ab253f6df16fe45a8b17fbd73c2
--- /dev/null
+++ b/test/tatoeba-test.ukr-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3a6382b72290b13ad2fc99f92e672c37e8f7553458d044abe50577ebcf3066d
+size 138693
diff --git a/test/tatoeba-test.vie-cmn_Hans.tsv b/test/tatoeba-test.vie-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..90749a57b18853c15d444e9cbfd3ac40e0cdbbbc
--- /dev/null
+++ b/test/tatoeba-test.vie-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5c26563b02beaa76b221dc754ddfdd228015eef16f324553ce2514188a4bc53
+size 38707
diff --git a/test/tatoeba-test.vie-vie.tsv b/test/tatoeba-test.vie-vie.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ea976f7112aeae30871c3bb855e18a6f7f76c784
--- /dev/null
+++ b/test/tatoeba-test.vie-vie.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2a0f2a570768d5cbb0be17b21e29bcc5c7cd57de7a3ee292639c672d45a523f
+size 68696
diff --git a/test/tatoeba-test.vie-zho.tsv b/test/tatoeba-test.vie-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..e1d092106e189bede6980de5f1e482400fea4191
--- /dev/null
+++ b/test/tatoeba-test.vie-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:38d778c4d22b6831661b38404bc704f7e7223f3c96aa7ac54fda2b5b30e7944c
+size 45707
diff --git a/test/tatoeba-test.wuu-cmn_Hans.tsv b/test/tatoeba-test.wuu-cmn_Hans.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..15254b7ab8fae992a3afbe433224dce8fd7f6df0
--- /dev/null
+++ b/test/tatoeba-test.wuu-cmn_Hans.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7db06802cffba741e2ef3da7ab66c2f65670efcf5ee1c4c09bbd4a02499e0503
+size 77430
diff --git a/test/tatoeba-test.yid-yid.tsv b/test/tatoeba-test.yid-yid.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..319e093f8e8194c68b61b58b74210fb83bea0699
--- /dev/null
+++ b/test/tatoeba-test.yid-yid.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11e24de00a30dd7d1d20afe28c430662075e2b99bf296aa51b8ddc6ca2082835
+size 35801
diff --git a/test/tatoeba-test.zho-zho.tsv b/test/tatoeba-test.zho-zho.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..29a1f6daecccb843f4a193691fc8333b071a35b5
--- /dev/null
+++ b/test/tatoeba-test.zho-zho.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24c9055e3149e259aa1f4c08e9265bdfde702278f561ea7be49982214403c00b
+size 264336
diff --git a/test/tatoeba-test.zsm_Latn-ind.tsv b/test/tatoeba-test.zsm_Latn-ind.tsv
new file mode 100644
index 0000000000000000000000000000000000000000..ec76323ebc946118b32e82103399e8fb72863cee
--- /dev/null
+++ b/test/tatoeba-test.zsm_Latn-ind.tsv
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:458ddb31eca986529d7e574efafd51b2a7a94e33ddda5b63aaba0198f395de15
+size 23540