Datasets:
ArXiv:
License:
update
Browse files- README.md +5 -5
- data/europa_eac_tm.jsonl +2 -2
- data/hind_encorp.jsonl +3 -0
- data/hrenwac_para.jsonl +3 -0
- dataset_details.md +219 -38
- docs/picture/europa_eac_tm_text_length.jpg +2 -2
- docs/picture/hind_encorp_text_length.jpg +3 -0
- docs/picture/hrenwac_para_text_length.jpg +3 -0
- examples/make_subset_details.py +1 -1
- examples/preprocess/preprocess_hind_encorp.py +85 -0
- examples/preprocess/preprocess_hrenwac_para.py +85 -0
- language_identification.py +5 -0
README.md
CHANGED
@@ -42,12 +42,12 @@ Tips:
|
|
42 |
| autshumato | | TRAIN: 652824 | Autshumato 项目的目标之一是开发三种南非语言对的机器翻译系统。 | [autshumato](https://huggingface.co/datasets/autshumato) |
|
43 |
| chr_en | [2010.04791](https://arxiv.org/abs/2010.04791) | 样本个数 | ChrEn 是切罗基语-英语并行数据集,用于促进切罗基语和英语之间的机器翻译研究。 ChrEn 资源极少,总共包含 14k 个句子对,其分割方式有利于域内和域外评估。 ChrEn 还包含 5k 切罗基语单语数据以实现半监督学习。 | [chr_en](https://huggingface.co/datasets/chr_en) |
|
44 |
| cmu_hinglish_dog | [CMU_DoG](https://github.com/festvox/datasets-CMU_DoG); [1809.07358](https://arxiv.org/abs/1809.07358) | TRAIN: 13146, VALID: 1645, TEST: 1616 | 这是印度英语(印地语-英语之间的代码混合)文本对话及其相应的英语版本的集合。 可用于两者之间的翻译。 该数据集由 CMU 的 Alan Black 教授团队提供。 | [cmu_hinglish_dog](https://huggingface.co/datasets/cmu_hinglish_dog) |
|
45 |
-
| europa_eac_tm | [EAC-Translation Memory](https://joint-research-centre.ec.europa.eu/language-technology-resources/eac-translation-memory_en) | TRAIN:
|
46 |
| europa_ecdc_tm | [ECDC-Translation Memory](https://joint-research-centre.ec.europa.eu/language-technology-resources/ecdc-translation-memory_en) | TRAIN: 58968 | 2012 年 10 月,欧盟 (EU) 机构“欧洲疾病预防和控制中心”(ECDC) 发布了翻译记忆库 (TM),即 25 种语言的句子及其专业翻译的集合。 | [europa_ecdc_tm](https://huggingface.co/datasets/europa_ecdc_tm) |
|
47 |
-
| flores | [1902.01382](https://arxiv.org/abs/1902.01382) |
|
48 |
-
| giga_fren | |
|
49 |
-
| hind_encorp | [HindEnCorp](https://aclanthology.org/L14-1643/) |
|
50 |
-
| hrenwac_para | |
|
51 |
| id_panl_bppt | | 样本个数 | BPPT(印度尼西亚技术评估和应用机构)为 PAN 本地化项目(发展亚洲本地语言计算能力的区域性倡议)创建的多域翻译系统并行文本语料库。 该数据集包含大约 24K 个句子,分为 4 个不同主题(经济、国际、科学技术和体育)。 | [id_panl_bppt](https://huggingface.co/datasets/id_panl_bppt) |
|
52 |
| igbo | [Igbo-English Machine Translation](https://arxiv.org/abs/2004.00648v1) | 样本个数 | 在这项工作中,我们讨论了为伊博语(尼日利亚三种主要语言之一)构建标准机器翻译基准数据集所做的努力。 | [igbo_english_machine_translation](https://huggingface.co/datasets/igbo_english_machine_translation) |
|
53 |
| menyo20k_mt | [menyo20k_mt](https://arxiv.org/abs/2103.08647v3) | 样本个数 | MENYO-20k 是一个多域并行数据集,其中的文本来自新闻文章、ted 演讲、电影文字记录、广播文字记录、科��文本以及其他由网络和专业翻译人员策划的短文。 | [menyo20k_mt](https://huggingface.co/datasets/menyo20k_mt) |
|
|
|
42 |
| autshumato | | TRAIN: 652824 | Autshumato 项目的目标之一是开发三种南非语言对的机器翻译系统。 | [autshumato](https://huggingface.co/datasets/autshumato) |
|
43 |
| chr_en | [2010.04791](https://arxiv.org/abs/2010.04791) | 样本个数 | ChrEn 是切罗基语-英语并行数据集,用于促进切罗基语和英语之间的机器翻译研究。 ChrEn 资源极少,总共包含 14k 个句子对,其分割方式有利于域内和域外评估。 ChrEn 还包含 5k 切罗基语单语数据以实现半监督学习。 | [chr_en](https://huggingface.co/datasets/chr_en) |
|
44 |
| cmu_hinglish_dog | [CMU_DoG](https://github.com/festvox/datasets-CMU_DoG); [1809.07358](https://arxiv.org/abs/1809.07358) | TRAIN: 13146, VALID: 1645, TEST: 1616 | 这是印度英语(印地语-英语之间的代码混合)文本对话及其相应的英语版本的集合。 可用于两者之间的翻译。 该数据集由 CMU 的 Alan Black 教授团队提供。 | [cmu_hinglish_dog](https://huggingface.co/datasets/cmu_hinglish_dog) |
|
45 |
+
| europa_eac_tm | [EAC-Translation Memory](https://joint-research-centre.ec.europa.eu/language-technology-resources/eac-translation-memory_en) | TRAIN: 38054 | 该数据集是从英语到多达 25 种语言的手动翻译的语料库,由欧盟教育和文化总局 (EAC) 于 2012 年发布。 | [europa_eac_tm](https://huggingface.co/datasets/europa_eac_tm) |
|
46 |
| europa_ecdc_tm | [ECDC-Translation Memory](https://joint-research-centre.ec.europa.eu/language-technology-resources/ecdc-translation-memory_en) | TRAIN: 58968 | 2012 年 10 月,欧盟 (EU) 机构“欧洲疾病预防和控制中心”(ECDC) 发布了翻译记忆库 (TM),即 25 种语言的句子及其专业翻译的集合。 | [europa_ecdc_tm](https://huggingface.co/datasets/europa_ecdc_tm) |
|
47 |
+
| flores | [1902.01382](https://arxiv.org/abs/1902.01382) | | 低资源机器翻译的评估数据集:尼泊尔语-英语和僧伽罗语-英语。 | [flores](https://huggingface.co/datasets/flores) |
|
48 |
+
| giga_fren | | | | [giga_fren](https://huggingface.co/datasets/giga_fren) |
|
49 |
+
| hind_encorp | [HindEnCorp](https://aclanthology.org/L14-1643/) | TRAIN: 445071 | HindEnCorp 并行文本(句子对齐)来自以下来源:Tides,其中包含主要取自新闻文章的 50K 句对。 该数据集最初是为 2002 年 DARPA-TIDES 惊喜语言竞赛收集的,后来在 IIIT 海得拉巴进行了完善,并提供给 ICON 2008 的 NLP 工具竞赛(Venkatapathy,2008)。 | [hind_encorp](https://huggingface.co/datasets/hind_encorp) |
|
50 |
+
| hrenwac_para | | TRAIN: 191946 | hrenWaC 语料库版本 2.0 由从克罗地亚 .hr 顶级域爬取的并行克罗地亚语-英语文本组成。 | [hrenwac_para](https://huggingface.co/datasets/hrenwac_para) |
|
51 |
| id_panl_bppt | | 样本个数 | BPPT(印度尼西亚技术评估和应用机构)为 PAN 本地化项目(发展亚洲本地语言计算能力的区域性倡议)创建的多域翻译系统并行文本语料库。 该数据集包含大约 24K 个句子,分为 4 个不同主题(经济、国际、科学技术和体育)。 | [id_panl_bppt](https://huggingface.co/datasets/id_panl_bppt) |
|
52 |
| igbo | [Igbo-English Machine Translation](https://arxiv.org/abs/2004.00648v1) | 样本个数 | 在这项工作中,我们讨论了为伊博语(尼日利亚三种主要语言之一)构建标准机器翻译基准数据集所做的努力。 | [igbo_english_machine_translation](https://huggingface.co/datasets/igbo_english_machine_translation) |
|
53 |
| menyo20k_mt | [menyo20k_mt](https://arxiv.org/abs/2103.08647v3) | 样本个数 | MENYO-20k 是一个多域并行数据集,其中的文本来自新闻文章、ted 演讲、电影文字记录、广播文字记录、科��文本以及其他由网络和专业翻译人员策划的短文。 | [menyo20k_mt](https://huggingface.co/datasets/menyo20k_mt) |
|
data/europa_eac_tm.jsonl
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dcdc26ac2cbad4a661c448cde2dcd0641861b548f9200cfd148f7348ef2fa2d9
|
3 |
+
size 4896643
|
data/hind_encorp.jsonl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e000880c6418e01f0ba145ca570be5e32fdd25730d154b09f1283173c882584c
|
3 |
+
size 92643407
|
data/hrenwac_para.jsonl
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:673bec80093a2aee4e59e3d3919e5168739d75e2e31e85341baee4c2c77a5848
|
3 |
+
size 43817114
|
dataset_details.md
CHANGED
@@ -297,52 +297,132 @@ en: 5966
|
|
297 |
|
298 |
```text
|
299 |
语种数量:
|
300 |
-
|
301 |
-
|
302 |
-
|
303 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
304 |
```
|
305 |
|
306 |
样本示例:
|
307 |
|
308 |
| 数据 | 语种 | 样本 |
|
309 |
| :---: | :---: | :---: |
|
310 |
-
| europa_eac_tm | bg |
|
311 |
-
| europa_eac_tm | bg |
|
312 |
-
| europa_eac_tm | bg |
|
313 |
-
| europa_eac_tm |
|
314 |
-
| europa_eac_tm |
|
315 |
-
| europa_eac_tm |
|
316 |
-
| europa_eac_tm |
|
317 |
-
| europa_eac_tm |
|
318 |
-
| europa_eac_tm |
|
319 |
-
| europa_eac_tm |
|
320 |
-
| europa_eac_tm |
|
321 |
-
| europa_eac_tm |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
322 |
|
323 |
<details>
|
324 |
<summary>文本长度</summary>
|
325 |
-
<pre><code>0-10:
|
326 |
-
10-20:
|
327 |
-
20-30:
|
328 |
-
30-40:
|
329 |
-
40-50:
|
330 |
-
50-60:
|
331 |
-
60-70:
|
332 |
-
70-80:
|
333 |
-
80-90:
|
334 |
-
90-100:
|
335 |
-
100-110:
|
336 |
-
110-120:
|
337 |
-
120-130:
|
338 |
-
130-140:
|
339 |
-
140-150:
|
340 |
-
150-160:
|
341 |
-
160-170:
|
342 |
-
170-180:
|
343 |
-
180-190:
|
344 |
-
190-200:
|
345 |
-
200-210:
|
346 |
</code></pre>
|
347 |
</details>
|
348 |
|
@@ -385,7 +465,6 @@ ga: 1331
|
|
385 |
|
386 |
样本示例:
|
387 |
|
388 |
-
|
389 |
| 数据 | 语种 | 样本 |
|
390 |
| :---: | :---: | :---: |
|
391 |
| europa_ecdc_tm | bg | Засега няма ваксина срещу хепатит С. |
|
@@ -495,6 +574,108 @@ ga: 1331
|
|
495 |
![europa_ecdc_tm_text_length.jpg](docs/picture/europa_ecdc_tm_text_length.jpg)
|
496 |
|
497 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
498 |
#### iwslt2017
|
499 |
以下都是 train 训练集的信息
|
500 |
|
|
|
297 |
|
298 |
```text
|
299 |
语种数量:
|
300 |
+
bg: 2535
|
301 |
+
da: 2339
|
302 |
+
cs: 2273
|
303 |
+
pl: 2267
|
304 |
+
lt: 2235
|
305 |
+
sl: 2231
|
306 |
+
lv: 2213
|
307 |
+
en: 2102
|
308 |
+
de: 1968
|
309 |
+
fr: 1962
|
310 |
+
sk: 1958
|
311 |
+
es: 1898
|
312 |
+
pt: 1730
|
313 |
+
sv: 1643
|
314 |
+
hu: 1390
|
315 |
+
el: 1367
|
316 |
+
et: 1029
|
317 |
+
it: 1021
|
318 |
+
mt: 971
|
319 |
+
nl: 931
|
320 |
+
is: 626
|
321 |
+
ro: 543
|
322 |
+
fi: 501
|
323 |
+
no: 321
|
324 |
```
|
325 |
|
326 |
样本示例:
|
327 |
|
328 |
| 数据 | 语种 | 样本 |
|
329 |
| :---: | :---: | :---: |
|
330 |
+
| europa_eac_tm | bg | Събития с широка публичност |
|
331 |
+
| europa_eac_tm | bg | Кашмирски |
|
332 |
+
| europa_eac_tm | bg | Специалисти, свързани с преподаването на предучилищното образование |
|
333 |
+
| europa_eac_tm | da | Evenements A Haute Vuisibilite |
|
334 |
+
| europa_eac_tm | da | Kashmiri |
|
335 |
+
| europa_eac_tm | da | Medarbejdere i førskoleundervisning |
|
336 |
+
| europa_eac_tm | sl | Močno prepoznavni dogodki |
|
337 |
+
| europa_eac_tm | sl | kašmirščina |
|
338 |
+
| europa_eac_tm | sl | Pomočniki vzgojiteljev predšolskih otrok |
|
339 |
+
| europa_eac_tm | cs | kašmírština |
|
340 |
+
| europa_eac_tm | cs | Techničtí pracovníci – učitelé předškolní výchovy |
|
341 |
+
| europa_eac_tm | cs | Studium všeobecného typu bez zaměření na určitý předmět |
|
342 |
+
| europa_eac_tm | el | Κασμίρ |
|
343 |
+
| europa_eac_tm | el | Διαρθρωτικά Δίκτυα ERASMUS |
|
344 |
+
| europa_eac_tm | el | Το σχέδιο αντανακλά μια κοινή ανησυχία για την Ευρωπαϊκή κοινωνία, όπως ο ρατσισμός, η ξενοφοβία και ο αντισημιτισμός, η χρήση εξαρτησιογόνων ουσιών... |
|
345 |
+
| europa_eac_tm | es | Cachemir |
|
346 |
+
| europa_eac_tm | es | Profesionales relacionados con la enseñanza pre-primaria |
|
347 |
+
| europa_eac_tm | es | Programas generales sin énfasis en ninguna materia especial |
|
348 |
+
| europa_eac_tm | fi | Kasmiri |
|
349 |
+
| europa_eac_tm | fi | Hanke heijastelee eurooppalaisille yhteisiä huolenaiheita kuten rasismia, muukalaisvihaa, antisemitismiä tai huumausaineiden käyttöä. |
|
350 |
+
| europa_eac_tm | fi | Heard ja McDonaldinsaaret |
|
351 |
+
| europa_eac_tm | hu | HIDE |
|
352 |
+
| europa_eac_tm | hu | ERASMUS Strukturális Hálózatok |
|
353 |
+
| europa_eac_tm | hu | A projekt tükrözi az európai társadalom valamely közös problémáját, például a fajgyűlöletet, az idegengyűlöletet, az antiszemitizmust, a kábítószerekkel való visszaélést |
|
354 |
+
| europa_eac_tm | lt | Kašmyrų |
|
355 |
+
| europa_eac_tm | lt | Ikimokyklinio ugdymo jaunesnieji specialistai |
|
356 |
+
| europa_eac_tm | lt | Bendrosios programos neišskiriant atskirų dalykų |
|
357 |
+
| europa_eac_tm | lv | Kašmiriešu |
|
358 |
+
| europa_eac_tm | lv | Pirmsskolas izglītības apmācību asociētie speciālisti |
|
359 |
+
| europa_eac_tm | lv | Vispārizglītojošās programmas, neuzsverot nekādus īpašus mācību priekšmetus |
|
360 |
+
| europa_eac_tm | mt | Kaxmiri |
|
361 |
+
| europa_eac_tm | mt | Il-proġett jirrifletti tħassib komuni għas-soċjetà Ewropea, bħar-razziżmu, ksenofobija u anti-semitiżmu, abbuż tad-droga... |
|
362 |
+
| europa_eac_tm | mt | Il-Gżejjer Heard u l-Gżejjer Mcdonald |
|
363 |
+
| europa_eac_tm | nl | Kasjmiri |
|
364 |
+
| europa_eac_tm | nl | Het project weerspiegelt een gemeenschappelijk belang in de Europese samenleving, zoals racisme, xenofobie en anti-semitisme, drugmisbruik... |
|
365 |
+
| europa_eac_tm | nl | Heard- en McDonaldeilanden |
|
366 |
+
| europa_eac_tm | pl | kaszmirski |
|
367 |
+
| europa_eac_tm | pl | Asystenci nauczycieli wychowania przedszkolnego |
|
368 |
+
| europa_eac_tm | pl | Programy ogólne bez nacisku na określony przedmiot |
|
369 |
+
| europa_eac_tm | de | Lehrberufe der Vorschulstufe |
|
370 |
+
| europa_eac_tm | de | Allgemeine Programme ohne Schwerpunkt auf speziellen Fächern |
|
371 |
+
| europa_eac_tm | de | ERASMUS strukturelle Netzwerke |
|
372 |
+
| europa_eac_tm | en | Pre-primary education teaching associate professionals |
|
373 |
+
| europa_eac_tm | en | General programmes with no special subject emphasis |
|
374 |
+
| europa_eac_tm | en | ERASMUS Structural Networks |
|
375 |
+
| europa_eac_tm | fr | Professionnels associés à l'enseignement pré-primaire (maternelle en France) |
|
376 |
+
| europa_eac_tm | fr | Programmes généraux sans thématique particulière |
|
377 |
+
| europa_eac_tm | fr | ERASMUS Réseaux structurels |
|
378 |
+
| europa_eac_tm | pt | Profissionais similares do ensino pré primário |
|
379 |
+
| europa_eac_tm | pt | Programas gerais sem ênfase em assuntos especiais |
|
380 |
+
| europa_eac_tm | pt | Redes Estruturais ERASMUS |
|
381 |
+
| europa_eac_tm | sk | Pomocní učitelia v predškolskej výchove |
|
382 |
+
| europa_eac_tm | sk | Všeobecnovzdelávacie programy bez osobitného dôrazu na predmet |
|
383 |
+
| europa_eac_tm | sk | ERASMUS Štrukturálne siete |
|
384 |
+
| europa_eac_tm | sv | Yrkesverksamma knutna till sektorn förskoleutbildning |
|
385 |
+
| europa_eac_tm | sv | Allmänna program utan särskild inriktning |
|
386 |
+
| europa_eac_tm | sv | ERASMUS strukturella nätverk |
|
387 |
+
| europa_eac_tm | et | Üldõppekavad ilma süvaõppeta |
|
388 |
+
| europa_eac_tm | et | Projekt peegeldab selliseid Euroopa ühiskonna murekohti nagu rassism, ksenofoobia, antisemitism, narkootikumide tarbimine jne |
|
389 |
+
| europa_eac_tm | et | Heard ja McDonald saared |
|
390 |
+
| europa_eac_tm | ro | Programe generale fară accent pe o anumită disciplină |
|
391 |
+
| europa_eac_tm | ro | ERASMUS Structural Network |
|
392 |
+
| europa_eac_tm | ro | Proiectul reflectă un subiect de preocupare comun pentru societatea europeană, precum rasismul, xenofobia şi antisemitismul, consumul de droguri, etc |
|
393 |
+
| europa_eac_tm | is | Verkefnið endurspeglar sameiginlegar áhyggjur fyrir íbúa Evrópu af málefnum svo sem mismunun eftir kynþætti, trú, ólíkri menningu, lyfjamisnotkun... |
|
394 |
+
| europa_eac_tm | is | Farandsamfélög (t.d. Sígaunar) |
|
395 |
+
| europa_eac_tm | is | Framleiðsla á hráefnum til lyfjaframleiðslu og lyfjaundirbúnings |
|
396 |
+
| europa_eac_tm | it | Il progetto riflette una preoccupazione comune per i problemi della società europea, come il razzismo, la xenofobia e l'antisemitismo, l'abuso di droghe, ecc. |
|
397 |
+
| europa_eac_tm | it | Isole Heard e Mcdonald |
|
398 |
+
| europa_eac_tm | it | Comunità Rom |
|
399 |
+
| europa_eac_tm | no | Prosjektet gjenspeiler et felles engasjement i europeiske samfunnsspørsmål, f.eks. rasisme, fremmedfrykt, antisemittisme eller narkotikamisbruk |
|
400 |
+
| europa_eac_tm | no | Heard Island og Mcdonald Islands |
|
401 |
+
| europa_eac_tm | no | Roma-samfunn |
|
402 |
|
403 |
<details>
|
404 |
<summary>文本长度</summary>
|
405 |
+
<pre><code>0-10: 6273
|
406 |
+
10-20: 8092
|
407 |
+
20-30: 6645
|
408 |
+
30-40: 4941
|
409 |
+
40-50: 3155
|
410 |
+
50-60: 2100
|
411 |
+
60-70: 1453
|
412 |
+
70-80: 1131
|
413 |
+
80-90: 821
|
414 |
+
90-100: 612
|
415 |
+
100-110: 468
|
416 |
+
110-120: 359
|
417 |
+
120-130: 245
|
418 |
+
130-140: 217
|
419 |
+
140-150: 201
|
420 |
+
150-160: 173
|
421 |
+
160-170: 146
|
422 |
+
170-180: 131
|
423 |
+
180-190: 92
|
424 |
+
190-200: 90
|
425 |
+
200-210: 709
|
426 |
</code></pre>
|
427 |
</details>
|
428 |
|
|
|
465 |
|
466 |
样本示例:
|
467 |
|
|
|
468 |
| 数据 | 语种 | 样本 |
|
469 |
| :---: | :---: | :---: |
|
470 |
| europa_ecdc_tm | bg | Засега няма ваксина срещу хепатит С. |
|
|
|
574 |
![europa_ecdc_tm_text_length.jpg](docs/picture/europa_ecdc_tm_text_length.jpg)
|
575 |
|
576 |
|
577 |
+
#### hind_encorp
|
578 |
+
以下都是 train 训练集的信息
|
579 |
+
|
580 |
+
```text
|
581 |
+
语种数量:
|
582 |
+
en: 229883
|
583 |
+
hi: 215188
|
584 |
+
```
|
585 |
+
|
586 |
+
样本示例:
|
587 |
+
|
588 |
+
| 数据 | 语种 | 样本 |
|
589 |
+
| :---: | :---: | :---: |
|
590 |
+
| hind_encorp | en | Sharaabi |
|
591 |
+
| hind_encorp | en | politicians do not have permission to do what needs to be done. |
|
592 |
+
| hind_encorp | en | I'd like to tell you about one such child, |
|
593 |
+
| hind_encorp | hi | शराबी |
|
594 |
+
| hind_encorp | hi | राजनीतिज्ञों के पास जो कार्य करना चाहिए, वह करने कि अनुमति नहीं है . |
|
595 |
+
| hind_encorp | hi | मई आपको ऐसे ही एक बच्चे के बारे में बताना चाहूंगी, |
|
596 |
+
|
597 |
+
<details>
|
598 |
+
<summary>文本长度</summary>
|
599 |
+
<pre><code>0-10: 30653
|
600 |
+
10-20: 73226
|
601 |
+
20-30: 53635
|
602 |
+
30-40: 44904
|
603 |
+
40-50: 37282
|
604 |
+
50-60: 29459
|
605 |
+
60-70: 22956
|
606 |
+
70-80: 18088
|
607 |
+
80-90: 15361
|
608 |
+
90-100: 13614
|
609 |
+
100-110: 12327
|
610 |
+
110-120: 11161
|
611 |
+
120-130: 10286
|
612 |
+
130-140: 8782
|
613 |
+
140-150: 7767
|
614 |
+
150-160: 6571
|
615 |
+
160-170: 5752
|
616 |
+
170-180: 4803
|
617 |
+
180-190: 4165
|
618 |
+
190-200: 3510
|
619 |
+
200-210: 30769
|
620 |
+
</code></pre>
|
621 |
+
</details>
|
622 |
+
|
623 |
+
文本长度统计图像:
|
624 |
+
|
625 |
+
![hind_encorp_text_length.jpg](docs/picture/hind_encorp_text_length.jpg)
|
626 |
+
|
627 |
+
|
628 |
+
#### hrenwac_para
|
629 |
+
以下都是 train 训练集的信息
|
630 |
+
|
631 |
+
```text
|
632 |
+
语种数量:
|
633 |
+
en: 96102
|
634 |
+
hr: 95844
|
635 |
+
```
|
636 |
+
|
637 |
+
样本示例:
|
638 |
+
|
639 |
+
| 数据 | 语种 | 样本 |
|
640 |
+
| :---: | :---: | :---: |
|
641 |
+
| hrenwac_para | en | There is probably no person in the world that the news about the medical operation on close someone does not feel sadness and helplessness. |
|
642 |
+
| hrenwac_para | en | After news of the black on weekends in the traffic is often not see the sequel in which we participate - and maybe save someone's life. |
|
643 |
+
| hrenwac_para | en | There is something that can't be bought, because they can't produce, but only to give. |
|
644 |
+
| hrenwac_para | hr | Vjerojatno ne postoji osoba na svijetu koja na vijest o medicinskom zahvatu nekoga bliskoga ne osjeća tugu i bespomoćnost. |
|
645 |
+
| hrenwac_para | hr | Iza vijesti o crnim vikendima u prometu često ne vidimo onaj nastavak u kojemu možemo sudjelovati – i možda spasiti nečiji život. |
|
646 |
+
| hrenwac_para | hr | Postoji nešto što se ne može kupiti, jer se ne može proizvesti, nego samo darovati. |
|
647 |
+
|
648 |
+
<details>
|
649 |
+
<summary>文本长度</summary>
|
650 |
+
<pre><code>0-10: 253
|
651 |
+
10-20: 1160
|
652 |
+
20-30: 2194
|
653 |
+
30-40: 4350
|
654 |
+
40-50: 5911
|
655 |
+
50-60: 7780
|
656 |
+
60-70: 9253
|
657 |
+
70-80: 10465
|
658 |
+
80-90: 11194
|
659 |
+
90-100: 11641
|
660 |
+
100-110: 11608
|
661 |
+
110-120: 11566
|
662 |
+
120-130: 10983
|
663 |
+
130-140: 10269
|
664 |
+
140-150: 9653
|
665 |
+
150-160: 8716
|
666 |
+
160-170: 7885
|
667 |
+
170-180: 6830
|
668 |
+
180-190: 6241
|
669 |
+
190-200: 5500
|
670 |
+
200-210: 38494
|
671 |
+
</code></pre>
|
672 |
+
</details>
|
673 |
+
|
674 |
+
文本长度统计图像:
|
675 |
+
|
676 |
+
![hrenwac_para_text_length.jpg](docs/picture/hrenwac_para_text_length.jpg)
|
677 |
+
|
678 |
+
|
679 |
#### iwslt2017
|
680 |
以下都是 train 训练集的信息
|
681 |
|
docs/picture/europa_eac_tm_text_length.jpg
CHANGED
Git LFS Details
|
Git LFS Details
|
docs/picture/hind_encorp_text_length.jpg
ADDED
Git LFS Details
|
docs/picture/hrenwac_para_text_length.jpg
ADDED
Git LFS Details
|
examples/make_subset_details.py
CHANGED
@@ -12,7 +12,7 @@ from project_settings import project_path
|
|
12 |
|
13 |
def get_args():
|
14 |
parser = argparse.ArgumentParser()
|
15 |
-
parser.add_argument("--dataset_name", default="
|
16 |
parser.add_argument(
|
17 |
"--dataset_cache_dir",
|
18 |
default=(project_path / "hub_datasets").as_posix(),
|
|
|
12 |
|
13 |
def get_args():
|
14 |
parser = argparse.ArgumentParser()
|
15 |
+
parser.add_argument("--dataset_name", default="hrenwac_para", type=str)
|
16 |
parser.add_argument(
|
17 |
"--dataset_cache_dir",
|
18 |
default=(project_path / "hub_datasets").as_posix(),
|
examples/preprocess/preprocess_hind_encorp.py
ADDED
@@ -0,0 +1,85 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/python3
|
2 |
+
# -*- coding: utf-8 -*-
|
3 |
+
import argparse
|
4 |
+
from collections import defaultdict
|
5 |
+
import json
|
6 |
+
import os
|
7 |
+
import sys
|
8 |
+
|
9 |
+
pwd = os.path.abspath(os.path.dirname(__file__))
|
10 |
+
sys.path.append(os.path.join(pwd, "../../"))
|
11 |
+
|
12 |
+
from datasets import load_dataset, DownloadMode
|
13 |
+
from tqdm import tqdm
|
14 |
+
|
15 |
+
from language_identification import LANGUAGE_MAP
|
16 |
+
from project_settings import project_path
|
17 |
+
|
18 |
+
|
19 |
+
def get_args():
|
20 |
+
parser = argparse.ArgumentParser()
|
21 |
+
parser.add_argument("--dataset_path", default="hind_encorp", type=str)
|
22 |
+
parser.add_argument(
|
23 |
+
"--dataset_cache_dir",
|
24 |
+
default=(project_path / "hub_datasets").as_posix(),
|
25 |
+
type=str
|
26 |
+
)
|
27 |
+
parser.add_argument(
|
28 |
+
"--output_file",
|
29 |
+
default=(project_path / "data/hind_encorp.jsonl"),
|
30 |
+
type=str
|
31 |
+
)
|
32 |
+
|
33 |
+
args = parser.parse_args()
|
34 |
+
return args
|
35 |
+
|
36 |
+
|
37 |
+
def main():
|
38 |
+
args = get_args()
|
39 |
+
|
40 |
+
dataset_dict = load_dataset(
|
41 |
+
path=args.dataset_path,
|
42 |
+
cache_dir=args.dataset_cache_dir,
|
43 |
+
# download_mode=DownloadMode.FORCE_REDOWNLOAD
|
44 |
+
)
|
45 |
+
print(dataset_dict)
|
46 |
+
|
47 |
+
text_set = set()
|
48 |
+
counter = defaultdict(int)
|
49 |
+
with open(args.output_file, "w", encoding="utf-8") as f:
|
50 |
+
for k, v in dataset_dict.items():
|
51 |
+
split = k
|
52 |
+
if split not in ("train", "validation", "test"):
|
53 |
+
print("skip split: {}".format(split))
|
54 |
+
continue
|
55 |
+
|
56 |
+
for sample in tqdm(v):
|
57 |
+
|
58 |
+
translation = sample["translation"]
|
59 |
+
for language, text in translation.items():
|
60 |
+
text = text.strip()
|
61 |
+
|
62 |
+
if text in text_set:
|
63 |
+
continue
|
64 |
+
text_set.add(text)
|
65 |
+
|
66 |
+
if language not in LANGUAGE_MAP.keys():
|
67 |
+
raise AssertionError("language: {}, text: {}".format(language, text))
|
68 |
+
|
69 |
+
row = {
|
70 |
+
"text": text,
|
71 |
+
"language": language,
|
72 |
+
"data_source": "hind_encorp",
|
73 |
+
"split": split
|
74 |
+
}
|
75 |
+
row = json.dumps(row, ensure_ascii=False)
|
76 |
+
f.write("{}\n".format(row))
|
77 |
+
counter[split] += 1
|
78 |
+
|
79 |
+
print("counter: {}".format(counter))
|
80 |
+
|
81 |
+
return
|
82 |
+
|
83 |
+
|
84 |
+
if __name__ == '__main__':
|
85 |
+
main()
|
examples/preprocess/preprocess_hrenwac_para.py
ADDED
@@ -0,0 +1,85 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/python3
|
2 |
+
# -*- coding: utf-8 -*-
|
3 |
+
import argparse
|
4 |
+
from collections import defaultdict
|
5 |
+
import json
|
6 |
+
import os
|
7 |
+
import sys
|
8 |
+
|
9 |
+
pwd = os.path.abspath(os.path.dirname(__file__))
|
10 |
+
sys.path.append(os.path.join(pwd, "../../"))
|
11 |
+
|
12 |
+
from datasets import load_dataset, DownloadMode
|
13 |
+
from tqdm import tqdm
|
14 |
+
|
15 |
+
from language_identification import LANGUAGE_MAP
|
16 |
+
from project_settings import project_path
|
17 |
+
|
18 |
+
|
19 |
+
def get_args():
|
20 |
+
parser = argparse.ArgumentParser()
|
21 |
+
parser.add_argument("--dataset_path", default="hrenwac_para", type=str)
|
22 |
+
parser.add_argument(
|
23 |
+
"--dataset_cache_dir",
|
24 |
+
default=(project_path / "hub_datasets").as_posix(),
|
25 |
+
type=str
|
26 |
+
)
|
27 |
+
parser.add_argument(
|
28 |
+
"--output_file",
|
29 |
+
default=(project_path / "data/hrenwac_para.jsonl"),
|
30 |
+
type=str
|
31 |
+
)
|
32 |
+
|
33 |
+
args = parser.parse_args()
|
34 |
+
return args
|
35 |
+
|
36 |
+
|
37 |
+
def main():
|
38 |
+
args = get_args()
|
39 |
+
|
40 |
+
dataset_dict = load_dataset(
|
41 |
+
path=args.dataset_path,
|
42 |
+
cache_dir=args.dataset_cache_dir,
|
43 |
+
# download_mode=DownloadMode.FORCE_REDOWNLOAD
|
44 |
+
)
|
45 |
+
print(dataset_dict)
|
46 |
+
|
47 |
+
text_set = set()
|
48 |
+
counter = defaultdict(int)
|
49 |
+
with open(args.output_file, "w", encoding="utf-8") as f:
|
50 |
+
for k, v in dataset_dict.items():
|
51 |
+
split = k
|
52 |
+
if split not in ("train", "validation", "test"):
|
53 |
+
print("skip split: {}".format(split))
|
54 |
+
continue
|
55 |
+
|
56 |
+
for sample in tqdm(v):
|
57 |
+
|
58 |
+
translation = sample["translation"]
|
59 |
+
for language, text in translation.items():
|
60 |
+
text = text.strip()
|
61 |
+
|
62 |
+
if text in text_set:
|
63 |
+
continue
|
64 |
+
text_set.add(text)
|
65 |
+
|
66 |
+
if language not in LANGUAGE_MAP.keys():
|
67 |
+
raise AssertionError("language: {}, text: {}".format(language, text))
|
68 |
+
|
69 |
+
row = {
|
70 |
+
"text": text,
|
71 |
+
"language": language,
|
72 |
+
"data_source": "hrenwac_para",
|
73 |
+
"split": split
|
74 |
+
}
|
75 |
+
row = json.dumps(row, ensure_ascii=False)
|
76 |
+
f.write("{}\n".format(row))
|
77 |
+
counter[split] += 1
|
78 |
+
|
79 |
+
print("counter: {}".format(counter))
|
80 |
+
|
81 |
+
return
|
82 |
+
|
83 |
+
|
84 |
+
if __name__ == '__main__':
|
85 |
+
main()
|
language_identification.py
CHANGED
@@ -16,6 +16,8 @@ _URLS = {
|
|
16 |
"cmu_hinglish_dog": "data/cmu_hinglish_dog.jsonl",
|
17 |
"europa_eac_tm": "data/europa_eac_tm.jsonl",
|
18 |
"europa_ecdc_tm": "data/europa_ecdc_tm.jsonl",
|
|
|
|
|
19 |
"iwslt2017": "data/iwslt2017.jsonl",
|
20 |
"mike0307": "data/mike0307.jsonl",
|
21 |
"nbnn": "data/nbnn.jsonl",
|
@@ -58,6 +60,7 @@ LANGUAGE_MAP = {
|
|
58 |
"gl": "galician",
|
59 |
"hi": "hindi",
|
60 |
"hi_en": "hindi english",
|
|
|
61 |
"hu": "hungarian",
|
62 |
"is": "icelandic",
|
63 |
"it": "italian",
|
@@ -103,6 +106,8 @@ class LanguageIdentification(datasets.GeneratorBasedBuilder):
|
|
103 |
datasets.BuilderConfig(name="cmu_hinglish_dog", version=VERSION, description="cmu_hinglish_dog"),
|
104 |
datasets.BuilderConfig(name="europa_eac_tm", version=VERSION, description="europa_eac_tm"),
|
105 |
datasets.BuilderConfig(name="europa_ecdc_tm", version=VERSION, description="europa_ecdc_tm"),
|
|
|
|
|
106 |
datasets.BuilderConfig(name="iwslt2017", version=VERSION, description="iwslt2017"),
|
107 |
datasets.BuilderConfig(name="mike0307", version=VERSION, description="mike0307"),
|
108 |
datasets.BuilderConfig(name="nbnn", version=VERSION, description="nbnn"),
|
|
|
16 |
"cmu_hinglish_dog": "data/cmu_hinglish_dog.jsonl",
|
17 |
"europa_eac_tm": "data/europa_eac_tm.jsonl",
|
18 |
"europa_ecdc_tm": "data/europa_ecdc_tm.jsonl",
|
19 |
+
"hind_encorp": "data/hind_encorp.jsonl",
|
20 |
+
"hrenwac_para": "data/hrenwac_para.jsonl",
|
21 |
"iwslt2017": "data/iwslt2017.jsonl",
|
22 |
"mike0307": "data/mike0307.jsonl",
|
23 |
"nbnn": "data/nbnn.jsonl",
|
|
|
60 |
"gl": "galician",
|
61 |
"hi": "hindi",
|
62 |
"hi_en": "hindi english",
|
63 |
+
"hr": "croatian",
|
64 |
"hu": "hungarian",
|
65 |
"is": "icelandic",
|
66 |
"it": "italian",
|
|
|
106 |
datasets.BuilderConfig(name="cmu_hinglish_dog", version=VERSION, description="cmu_hinglish_dog"),
|
107 |
datasets.BuilderConfig(name="europa_eac_tm", version=VERSION, description="europa_eac_tm"),
|
108 |
datasets.BuilderConfig(name="europa_ecdc_tm", version=VERSION, description="europa_ecdc_tm"),
|
109 |
+
datasets.BuilderConfig(name="hind_encorp", version=VERSION, description="hind_encorp"),
|
110 |
+
datasets.BuilderConfig(name="hrenwac_para", version=VERSION, description="hrenwac_para"),
|
111 |
datasets.BuilderConfig(name="iwslt2017", version=VERSION, description="iwslt2017"),
|
112 |
datasets.BuilderConfig(name="mike0307", version=VERSION, description="mike0307"),
|
113 |
datasets.BuilderConfig(name="nbnn", version=VERSION, description="nbnn"),
|