update

README.md

@@ -12,6 +12,7 @@ Tips:
 数据集从网上收集整理如下:
 
 
+
 多语言语料
 
 | 数据 | 原始数据/项目地址 | 样本个数 | 原始数据描述 | 替代数据下载地址 |
@@ -21,6 +22,7 @@ Tips:
 | stsb_multi_mt | [SemEval-2017 Task 1](https://arxiv.org/abs/1708.00055) | TRAIN: 104117, VALID: 25943, TEST: 22457 | **使用时注意要打乱**。可用语言有：de、en、es、fr、it、nl、pl、pt、ru、zh | [stsb_multi_mt](https://huggingface.co/datasets/stsb_multi_mt) |
 
 
+
 语种识别
 
 | 数据 | 原始数据/项目地址 | 样本个数 | 原始数据描述 | 替代数据下载地址 |
@@ -31,6 +33,7 @@ Tips:
 | nbnn | [oai-nb-no-sbr-80](https://www.nb.no/sprakbanken/ressurskatalog/oai-nb-no-sbr-80/) | TRAIN: 1556212, VALID: 1957, TEST: 1944 | 该语料库包含挪威电报局 (NTB) 的新闻文本从博克马尔语翻译成新挪威语的内容。 | [NbAiLab/nbnn_language_detection](https://huggingface.co/datasets/NbAiLab/nbnn_language_detection) |
 
 
+
 机器翻译
 
 | 数据 | 原始数据/项目地址 | 样本个数 | 原始数据描述 | 替代数据下载地址 |
@@ -50,13 +53,38 @@ Tips:
 | id_panl_bppt |  | TRAIN: 47916 | 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) |
 | igbo | [Igbo-English Machine Translation](https://arxiv.org/abs/2004.00648v1) |  | 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) |
 | menyo20k_mt | [menyo20k_mt](https://arxiv.org/abs/2103.08647v3) | TRAIN: 19899, VALID: 6655, TEST: 13148 | 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、科技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) |
-| para_pat | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 10242500 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
 | pib | [CVIT-PIB](https://arxiv.org/abs/2008.04860) |  | 该数据集是 11 种印度语言的大规模句子对齐语料库，即： CVIT-PIB 语料库是印度语言可用的最大多语言语料库。 | [pib](https://huggingface.co/datasets/pib) |
 | poleval2019_mt |  |  | PolEval 是一项受 SemEval 启发的波兰语自然语言处理工具评估活动。 | [poleval2019_mt](https://huggingface.co/datasets/poleval2019_mt) |
 | wmt19 | [statmt.org](https://www.statmt.org/wmt19/translation-task.html) | 样本个数 | 我们的目标是尽可能使用公开的数据源。我们的训练数据主要来源是Europarl 语料库、 UN 语料库、新闻评论语料库和 ParaCrawl语料库。我们还发布了单语 新闻抓取语料库。将提供其他特定语言的语料库。 | [wmt/wmt19](https://huggingface.co/datasets/wmt/wmt19) |
 | ro_sts_parallel |  | 样本个数 | 我们提出 RO-STS-Parallel - 通过将 STS 英语数据集翻译成罗马尼亚语而获得的并行罗马尼亚语-英语数据集。 | [ro_sts_parallel](https://huggingface.co/datasets/ro_sts_parallel) |
 
 
+
+机器翻译
+
+| 数据 | 原始数据/项目地址 | 样本个数 | 原始数据描述 | 替代数据下载地址 |
+| :--- |    :---:      |   :---: |    :---:    |     :---:     |
+| para_pat_cs_en | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 156028 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_de_en | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 3065565 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_de_fr | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 1243643 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_el_en | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 20234 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_en_es | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 1147278 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_en_hu | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 84824 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_en_ja | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 11971591 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_en_ko | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 4268110 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_en_pt | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 42623 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_en_ro | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 94326 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_en_ru | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 6795724 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_en_sk | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 44337 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_en_uk | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 177043 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_en_zh | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 9367823 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_es_fr | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 55795 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_fr_ja | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 599299 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_fr_ko | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 200044 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+| para_pat_fr_ru | [ParaPat](https://aclanthology.org/2020.lrec-1.465.pdf); [Homepage](https://figshare.com/articles/dataset/ParaPat_The_Multi-Million_Sentences_Parallel_Corpus_of_Patents_Abstracts/12627632) | TRAIN: 19577 | ParaPat：专利摘要的数百万个句子平行语料库 | [para_pat](https://huggingface.co/datasets/para_pat) |
+
+
+
 机器翻译
 
 https://opus.nlpl.eu/
@@ -78,6 +106,7 @@ https://opus.nlpl.eu/
 | tanzil | [Tanzil](https://opus.nlpl.eu/Tanzil/corpus/version/Tanzil) | 样本个数 |  | [tanzil](https://huggingface.co/datasets/tanzil) |
 
 
+
 机器翻译
 
 https://opus.nlpl.eu/
@@ -103,6 +132,9 @@ https://opus.nlpl.eu/
 | para_crawl_en_pl | [ParaCrawl](https://opus.nlpl.eu/ParaCrawl/corpus/version/ParaCrawl); [ParaCrawl](https://aclanthology.org/2020.acl-main.417.pdf) | TRAIN: 6537110 | 欧洲官方语言的网络规模并行语料库。 | [para_crawl](https://huggingface.co/datasets/para_crawl) |
 | para_crawl_en_pt | [ParaCrawl](https://opus.nlpl.eu/ParaCrawl/corpus/version/ParaCrawl); [ParaCrawl](https://aclanthology.org/2020.acl-main.417.pdf) | TRAIN: 15186124 | 欧洲官方语言的网络规模并行语料库。 | [para_crawl](https://huggingface.co/datasets/para_crawl) |
 | para_crawl_en_ro | [ParaCrawl](https://opus.nlpl.eu/ParaCrawl/corpus/version/ParaCrawl); [ParaCrawl](https://aclanthology.org/2020.acl-main.417.pdf) | TRAIN: 3580912 | 欧洲官方语言的网络规模并行语料库。 | [para_crawl](https://huggingface.co/datasets/para_crawl) |
+| para_crawl_en_sk | [ParaCrawl](https://opus.nlpl.eu/ParaCrawl/corpus/version/ParaCrawl); [ParaCrawl](https://aclanthology.org/2020.acl-main.417.pdf) | TRAIN: 3047345 | 欧洲官方语言的网络规模并行语料库。 | [para_crawl](https://huggingface.co/datasets/para_crawl) |
+| para_crawl_en_sl | [ParaCrawl](https://opus.nlpl.eu/ParaCrawl/corpus/version/ParaCrawl); [ParaCrawl](https://aclanthology.org/2020.acl-main.417.pdf) | TRAIN: 1282153 | 欧洲官方语言的网络规模并行语料库。 | [para_crawl](https://huggingface.co/datasets/para_crawl) |
+| para_crawl_en_sv | [ParaCrawl](https://opus.nlpl.eu/ParaCrawl/corpus/version/ParaCrawl); [ParaCrawl](https://aclanthology.org/2020.acl-main.417.pdf) | TRAIN: 6626302 | 欧洲官方语言的网络规模并行语料库。 | [para_crawl](https://huggingface.co/datasets/para_crawl) |
 
 
 

data/{para_pat.jsonl → para_crawl_en_sk.jsonl}

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:125ac7672e3164064d3512f359a884150ae9e9e59f37622f0572157407fb689d
-size 13516036688
+oid sha256:a87fc26cb81e89f6a358ff19af120999c8d0a84b41340e4b53692d3bb7541d60
+size 552364035

data/para_crawl_en_sl.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0448d0a8b45908902c2f9a5eba3449bab096f700dda831aadaaa77f03729d08e
+size 273962941

data/para_crawl_en_sv.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2517d9126dd4ea6a77af63f32b7868a8aa2bbcf46de1df359dbb40548f666fea
+size 1335220345

data/para_pat_cs_en.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef48b64bd626dfecf3664d2197081e9af59e928c367b3eb106c82c66226051c6
+size 126803615

data/para_pat_de_en.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a8bf2cd12c8398eb5c8175c39775a3270a937c33248545c2e581c37e9df24ff
+size 980669225

data/para_pat_de_fr.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e6b42338438571675c804a64e385d07601178ad023d57615ebbcff7c675451a
+size 448168141

data/para_pat_el_en.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:262ae50b48f6969172360b0586691155794895884493a383e14e676c8411bbe6
+size 24704396

data/para_pat_en_es.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b766e53f9556bc49f30ca50a2f46c4b281fb998b95bf304faaa103837c9fbdb
+size 386858974

data/para_pat_en_hu.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f59210b7c6ca8c4ae40aeb215f50b88529552b2439b61359710f2bdb38695d7
+size 86090965

data/para_pat_en_ja.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d99a0d51b474a7ddbe56457bfd32737d8b7d8725b2480649060c8cc9b18ddc8
+size 3923552393

data/para_pat_en_ko.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15a2ac085c2fc7b02db098881001723f78aa449a285a7fd186f5d5a0eb89b2ae
+size 1654173028

data/para_pat_en_pt.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:62e7ffd26069282570dfcec22abc29ec9c0f1ea0b406e2dbb7736df6028126cd
+size 37598599

data/para_pat_en_ro.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64cdef5e751e2eecee7647a69aa4ba9f65dd831dff562ca53f60e2bddea99d0a
+size 83873351

data/para_pat_en_ru.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d0badc949e26d8fc705af2e590bc612b1fd01763ce8e9bd82820e53b7d34aa5
+size 2361506439

data/para_pat_en_sk.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f407a8a26a702889528798b70df77a794cda8652b168185b6c6b3eaadc0f60a6
+size 33038222

data/para_pat_en_uk.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89cf6760cd50ad81a6c73fcade9874211033daf7b08cfdb6e02b9262220b55dd
+size 148040587

data/para_pat_en_zh.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b93683a7e49ee2916b9d8292127dd2840e2e2de91727496189c7d2f2988b4e54
+size 2897727275

data/para_pat_es_fr.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99ebbbb2c82b33df8bf0ad7cf52bb59f1d5525cfa70d3596bfa6414fa048d8db
+size 49823412

data/para_pat_fr_ja.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab013f6c3f5d3638104afa47241663b45c74f9dd5801a572c87fd882b76edc7a
+size 246281219

data/para_pat_fr_ko.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8102b99d887072fa16c8b1e55cd939c8d089ad0a4eed74c8b87effeee3e5301
+size 197829003

data/para_pat_fr_ru.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35955e5097c0a74e3f92a1cc15607d9f0b1c90b00da286fa7680c1bd44e4dff1
+size 31750767

dataset_details.md

@@ -1535,6 +1535,211 @@ fr: 656
 ![open_subtitles_text_length.jpg](docs/picture/open_subtitles_text_length.jpg)
 
 
+#### para_crawl_en_pl
+以下都是 train 训练集的信息
+
+```text
+语种数量：
+en: 3268977
+pl: 3268133
+```
+
+样本示例：
+
+| 数据 | 语种 | 样本 |
+| :---: | :---: | :---: |
+| para_crawl_en_pl | en | 4. errors deltaE white – distortion palette of gray , analogous to the color deltaE |
+| para_crawl_en_pl | en | 4. errors deltaE white – distortion palette of gray, analogous to the color deltaE |
+| para_crawl_en_pl | en | A wy , also met with “ by professional test ” image quality at different hdmi cables ? Czekam na odpowiedzi w komentarzach |
+| para_crawl_en_pl | pl | 4. błędy deltaE bieli – przekłamania palety szarości, analogiczne do deltaE kolorów |
+| para_crawl_en_pl | pl | A wy, też spotkaliście się z „fachowymi testami” jakości obrazu na różnych kablach hdmi? Czekam na odpowiedzi w komentarzach |
+| para_crawl_en_pl | pl | Anatomia prawej terroru w Niemczech i niekompetentny BND tak raczej : "Homeland Security ??? " |
+
+
+<details>
+<summary>文本长度</summary>
+<pre><code>10-20: 13466
+20-30: 265266
+30-40: 495840
+40-50: 507676
+50-60: 532475
+60-70: 509995
+70-80: 455718
+80-90: 389501
+90-100: 358157
+100-110: 321469
+110-120: 281841
+120-130: 269854
+130-140: 263661
+140-150: 228062
+150-160: 196223
+160-170: 158115
+170-180: 138006
+180-190: 119215
+190-200: 107041
+200-210: 925529
+</code></pre>
+</details>
+
+文本长度统计图像：
+
+![para_crawl_en_pl_text_length.jpg](docs/picture/para_crawl_en_pl_text_length.jpg)
+
+
+#### para_crawl_en_pt
+以下都是 train 训练集的信息
+
+```text
+语种数量：
+en: 7604199
+pt: 7581925
+```
+
+样本示例：
+
+| 数据 | 语种 | 样本 |
+| :---: | :---: | :---: |
+| para_crawl_en_pt | en | 23 April 1905: Official ceremony the laying of the foundation stone . |
+| para_crawl_en_pt | en | 23 April 1905: Official ceremony the laying of the foundation stone. |
+| para_crawl_en_pt | en | Look familiar this face? Mmm….maybe not, but should be familiar to all of you. |
+| para_crawl_en_pt | pt | 23 Abril 1905: Cerimônia oficial colocação da primeira pedra . |
+| para_crawl_en_pt | pt | 23 Abril 1905: Cerimônia oficial colocação da primeira pedra. |
+| para_crawl_en_pt | pt | Look familiar this face? mmm….talvez não, mas deve ser familiar a todos vocês. |
+
+
+<details>
+<summary>文本长度</summary>
+<pre><code>10-20: 28282
+20-30: 493824
+30-40: 1007643
+40-50: 1171581
+50-60: 1330254
+60-70: 1238647
+70-80: 1050363
+80-90: 916893
+90-100: 817676
+100-110: 736767
+110-120: 651559
+120-130: 582723
+130-140: 525997
+140-150: 476314
+150-160: 441785
+160-170: 389696
+170-180: 352789
+180-190: 315622
+190-200: 288462
+200-210: 2369247
+</code></pre>
+</details>
+
+文本长度统计图像：
+
+![para_crawl_en_pt_text_length.jpg](docs/picture/para_crawl_en_pt_text_length.jpg)
+
+
+#### para_crawl_en_sl
+以下都是 train 训练集的信息
+
+```text
+语种数量：
+sl: 641207
+en: 640946
+```
+
+样本示例：
+
+| 数据 | 语种 | 样本 |
+| :---: | :---: | :---: |
+| para_crawl_en_sl | en | 1. First, press the START (or sign in the left corner of the screen) |
+| para_crawl_en_sl | en | An anatomy of the right terror in Germany and the incompetent BND so rather : "Homeland Security ??? ” |
+| para_crawl_en_sl | en | An anatomy of the right terror in Germany and the incompetent BND so rather: "Homeland Security ???” |
+| para_crawl_en_sl | sl | 1. Najprej pritisnite tipko START (ali znak v levem kotu zaslona) |
+| para_crawl_en_sl | sl | Anatomija pravi teror v Nemčiji in nesposobni BND tako precej : "Homeland Security ??? " |
+| para_crawl_en_sl | sl | Anatomija pravi teror v Nemčiji in nesposobni BND tako precej: "Homeland Security ???" |
+
+
+<details>
+<summary>文本长度</summary>
+<pre><code>10-20: 1135
+20-30: 33844
+30-40: 67979
+40-50: 78931
+50-60: 84083
+60-70: 83616
+70-80: 82419
+80-90: 80251
+90-100: 74910
+100-110: 65987
+110-120: 60489
+120-130: 56953
+130-140: 51168
+140-150: 47274
+150-160: 42314
+160-170: 38112
+170-180: 34451
+180-190: 32783
+190-200: 29340
+200-210: 236114
+</code></pre>
+</details>
+
+文本长度统计图像：
+
+![para_crawl_en_sl_text_length.jpg](docs/picture/para_crawl_en_sl_text_length.jpg)
+
+
+#### para_pat_en_uk
+以下都是 train 训练集的信息
+
+```text
+语种数量：
+uk: 88533
+en: 88510
+```
+
+样本示例：
+
+| 数据 | 语种 | 样本 |
+| :---: | :---: | :---: |
+| para_pat_en_uk | en | A replaceable handle to kitchen appliances comprises a bakelite handle with a connecting mechanism, available therein, a plastic part, which includes the upper section and the lower one, a spring, an aluminium part, which includes the upper section and the lower one. |
+| para_pat_en_uk | en | A method for predicting the risk of osteoporosis in the patients with systemic lupus erythematosus comprises X-ray imaging, the analysis of MTHFR C667T and eNOS T786C gene polymorphisms. The combination of the polymorphisms suggests the risk of osteoporosis. |
+| para_pat_en_uk | en | A method for growing red cabbage using the EM-preparation includes treating the soil with the given preparation prior to sowing with a rate of 20 l/ha. The seeds are soaked with a rate of 1 l/t and foliar fertilizings are carried out during vegetation with a rate of 2 l/ha in three terms. |
+| para_pat_en_uk | uk | Знімна ручка до кухонного приладдя містить бакелітову ручку з наявним у ній з'єднувальним механізмом, пластикову частину, яка включає верхню секцію і нижню секцію, пружину, алюмінієву частину, яка включає верхню секцію і нижню секцію. |
+| para_pat_en_uk | uk | Спосіб прогнозування розвитку остеопорозу при системному червоному вовчаку включає проведення рентгенографії, визначення поліморфізму генів MTHFR С667Т та eNOS Т786С, і при їх поєднанні прогнозування розвитку остеопорозу. |
+| para_pat_en_uk | uk | Спосіб вирощування капусти червоноголової з застосуванням ЕМ-препарату включає обробку даним препаратом ґрунту до посіву з нормою 20 л/га. Намочують насіння з нормою 1 л/т та здійснюють позакореневі підживлення під час вегетації з нормою 2 л/га в три строки. |
+
+
+<details>
+<summary>文本长度</summary>
+<pre><code>0-10: 10
+10-20: 4
+20-30: 2
+30-40: 15
+40-50: 37
+50-60: 63
+60-70: 133
+70-80: 248
+80-90: 400
+90-100: 555
+100-110: 754
+110-120: 923
+120-130: 1132
+130-140: 1278
+140-150: 1535
+150-160: 1685
+160-170: 1901
+170-180: 1998
+180-190: 2154
+190-200: 2404
+200-210: 159812
+</code></pre>
+</details>
+
+文本长度统计图像：
+
+![para_pat_en_uk_text_length.jpg](docs/picture/para_pat_en_uk_text_length.jpg)
+
+
 #### php
 以下都是 train 训练集的信息
 

examples/load_data/plan_1.py

@@ -1,17 +1,13 @@
 #!/usr/bin/python3
 # -*- coding: utf-8 -*-
-"""
-训练模型后发现它对短句子的识别能力很差。
-当句子长度足够的时候，它能够识别准确，并给出接近1.0 的概率值。
-但对于短句子，识别结果让人难以接受。例如把“你好”识别成 de 德语。
-"""
 import argparse
 import json
 import os
+import random
 import sys
 
 pwd = os.path.abspath(os.path.dirname(__file__))
-sys.path.append(os.path.join(pwd, "../../"))
+sys.path.append(os.path.join(pwd, "../../../"))
 
 from datasets import load_dataset, DownloadMode
 
@@ -114,14 +110,13 @@ def main():
             total = int(row[2])
             subsets = [e.strip() for e in row[3].split(";")]
 
-            train_count = 0
-            valid_count = 0
+            count = 0
             for subset in subsets:
                 if subset in subset_dataset_dict.keys():
                     dataset_dict = subset_dataset_dict[subset]
                 else:
                     dataset_dict = load_dataset(
-                        "../../language_identification.py",
+                        "qgyd2021/language_identification",
                         name=subset,
                         cache_dir=args.dataset_cache_dir,
                         # download_mode=DownloadMode.FORCE_REDOWNLOAD
@@ -134,38 +129,31 @@ def main():
                     language = sample["language"]
                     data_source = sample["data_source"]
 
-                    if train_count > total:
+                    if count > total:
                         break
-                    if language == abbr:
-                        row_ = {
-                            "text": text,
-                            "label": language,
-                            "data_source": data_source,
-                            "split": "train",
-                        }
-                        row_ = json.dumps(row_, ensure_ascii=False)
+
+                    if language != abbr:
+                        continue
+
+                    split = "train" if random.random() < 0.8 else "valid"
+
+                    row_ = {
+                        "text": text,
+                        "label": language,
+                        "language": full,
+                        "data_source": data_source,
+                        "split": split,
+                    }
+                    row_ = json.dumps(row_, ensure_ascii=False)
+
+                    if split == "train":
                         ftrain.write("{}\n".format(row_))
-                        train_count += 1
-
-                if "validation" in dataset_dict:
-                    valid_dataset = dataset_dict["validation"]
-                    for sample in valid_dataset:
-                        text = sample["text"]
-                        language = sample["language"]
-                        data_source = sample["data_source"]
-
-                        if valid_count > total:
-                            break
-                        if language == abbr:
-                            row_ = {
-                                "text": text,
-                                "label": language,
-                                "data_source": data_source,
-                                "split": "valid",
-                            }
-                            row_ = json.dumps(row_, ensure_ascii=False)
-                            fvalid.write("{}\n".format(row_))
-                            valid_count += 1
+                    elif split == "valid":
+                        fvalid.write("{}\n".format(row_))
+                    else:
+                        raise AssertionError
+
+                    count += 1
 
     return
 

examples/load_data/plan_2.py

@@ -0,0 +1,163 @@
+#!/usr/bin/python3
+# -*- coding: utf-8 -*-
+import argparse
+import json
+import os
+import random
+import sys
+
+pwd = os.path.abspath(os.path.dirname(__file__))
+sys.path.append(os.path.join(pwd, "../../"))
+
+from datasets import load_dataset, DownloadMode
+
+from project_settings import project_path
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--dataset_cache_dir",
+        default=(project_path / "hub_datasets").as_posix(),
+        type=str
+    )
+    parser.add_argument(
+        "--train_subset",
+        default="train.jsonl",
+        type=str
+    )
+    parser.add_argument(
+        "--valid_subset",
+        default="valid.jsonl",
+        type=str
+    )
+    args = parser.parse_args()
+    return args
+
+
+s = """
+|   ar   |     arabic     |  100000   |     iwslt2017    |
+|   bg   |   bulgarian    |  100000   |       xnli       |
+|   bn   |    bengali     |  36064    |  open_subtitles  |
+|   bs   |    bosnian     |  10212    |  open_subtitles  |
+|   cs   |     czech      |  100000   |       emea       |
+|   da   |     danish     |  100000   |  open_subtitles  |
+|   de   |     german     |  100000   |     iwslt2017    |
+|   el   |  modern greek  |  100000   |       emea       |
+|   en   |    english     |  100000   |     iwslt2017    |
+|   eo   |   esperanto    |  94101    | tatoeba; open_subtitles |
+|   es   |    spanish     |  100000   |       xnli       |
+|   et   |    estonian    |  100000   |       emea       |
+|   fi   |    finnish     |  100000   |    ecb; kde4     |
+|   fo   |    faroese     |  23807    |  nordic_langid   |
+|   fr   |    french      |  100000   |     iwslt2017    |
+|   ga   |    irish       |  100000   | multi_para_crawl |
+|   gl   |    galician    |  3096     |     tatoeba      |
+|   hi   |     hindi      |  100000   |       xnli       |
+|  hi_en |     hindi      |  7180     | cmu_hinglish_dog |
+|   hr   |    croatian    |  95844    |   hrenwac_para   |
+|   hu   |    hungarian   |  3801     |   europa_ecdc_tm; europa_eac_tm   |
+|   hy   |    armenian    |   660     |  open_subtitles  |
+|   id   |   indonesian   |  23940    |   id_panl_bppt   |
+|   is   |   icelandic    |  100000   | multi_para_crawl |
+|   it   |    italian     |  100000   |     iwslt2017    |
+|   ja   |    japanese    |  100000   |     iwslt2017    |
+|   ko   |    korean      |  100000   |     iwslt2017    |
+|   lt   |   lithuanian   |  100000   |       emea       |
+|   lv   |    latvian     |  100000   | multi_para_crawl |
+|   mr   |    marathi     |  51807    |     tatoeba      |
+|   mt   |    maltese     |  100000   | multi_para_crawl |
+|   nl   |    dutch       |  100000   |       kde4       |
+|   no   |   norwegian    |  100000   | multi_para_crawl |
+|   pl   |    polish      |  100000   | para_crawl_en_pl |
+|   pt   |   portuguese   |  100000   | para_crawl_en_pt |
+|   ro   |    romanian    |  100000   |     iwslt2017    |
+|   ru   |    russian     |  100000   |       xnli       |
+|   sk   |    slovak      |  100000   | multi_para_crawl |
+|   sl   |   slovenian    |  100000   | para_crawl_en_sl |
+|   sw   |    swahili     |  100000   |       xnli       |
+|   sv   |    swedish     |  100000   |       kde4       |
+|   th   |     thai       |  100000   |       xnli       |
+|   tl   |    tagalog     |  97241    | multi_para_crawl |
+|   tn   |    serpeti     |  100000   |   autshumato     |
+|   tr   |    turkish     |  100000   |       xnli       |
+|   ts   |    dzonga      |  100000   |    autshumato    |
+|   uk   |    ukrainian   |  88533    |  para_pat_en_uk  |
+|   ur   |     urdu       |  100000   |       xnli       |
+|   vi   |   vietnamese   |  100000   |       xnli       |
+|   yo   |     yoruba     |  9970    |    menyo20k_mt   |
+|   zh   |    chinese     |  100000   |       xnli       |
+|   zu   |  zulu, south africa  |  26801   |    autshumato    |
+"""
+
+
+def main():
+    args = get_args()
+
+    subset_dataset_dict = dict()
+
+    lines = s.strip().split("\n")
+
+    with open(args.train_subset, "w", encoding="utf-8") as ftrain, open(args.valid_subset, "w", encoding="utf-8") as fvalid:
+        for line in lines:
+            row = str(line).split("|")
+            row = [col.strip() for col in row if len(col) != 0]
+
+            if len(row) != 4:
+                raise AssertionError("not 4 item, line: {}".format(line))
+
+            abbr = row[0]
+            full = row[1]
+            total = int(row[2])
+            subsets = [e.strip() for e in row[3].split(";")]
+
+            count = 0
+            for subset in subsets:
+                if subset in subset_dataset_dict.keys():
+                    dataset_dict = subset_dataset_dict[subset]
+                else:
+                    dataset_dict = load_dataset(
+                        "qgyd2021/language_identification",
+                        name=subset,
+                        cache_dir=args.dataset_cache_dir,
+                        # download_mode=DownloadMode.FORCE_REDOWNLOAD
+                    )
+                    subset_dataset_dict[subset] = dataset_dict
+
+                train_dataset = dataset_dict["train"]
+                for sample in train_dataset:
+                    text = sample["text"]
+                    language = sample["language"]
+                    data_source = sample["data_source"]
+
+                    if count > total:
+                        break
+
+                    if language != abbr:
+                        continue
+
+                    split = "train" if random.random() < 0.8 else "valid"
+
+                    row_ = {
+                        "text": text,
+                        "label": language,
+                        "language": full,
+                        "data_source": data_source,
+                        "split": split,
+                    }
+                    row_ = json.dumps(row_, ensure_ascii=False)
+
+                    if split == "train":
+                        ftrain.write("{}\n".format(row_))
+                    elif split == "valid":
+                        fvalid.write("{}\n".format(row_))
+                    else:
+                        raise AssertionError
+
+                    count += 1
+
+    return
+
+
+if __name__ == "__main__":
+    main()

examples/make_subset_details.py

@@ -12,7 +12,7 @@ from project_settings import project_path
 
 def get_args():
     parser = argparse.ArgumentParser()
-    parser.add_argument("--dataset_name", default="tatoeba", type=str)
+    parser.add_argument("--dataset_name", default="para_pat_en_uk", type=str)
     parser.add_argument(
         "--dataset_cache_dir",
         default=(project_path / "hub_datasets").as_posix(),

examples/preprocess/preprocess_para_crawl.py

@@ -27,7 +27,7 @@ def get_args():
     )
     parser.add_argument(
         "--output_file",
-        default=(project_path / "data/para_crawl_en_ro.jsonl"),
+        default=(project_path / "data/para_crawl_en_sv.jsonl"),
         type=str
     )
 
@@ -58,12 +58,12 @@ def main():
         # "ennl",
         # "enpl",
         # "enpt",
-        "enro",
+        # "enro",
         # "ensk",
-        # "ensl", "ensv"
+        # "ensl",
+        "ensv"
     ]
 
-    # TODO: 数据集太大，加载不完。
     text_set = set()
     counter = defaultdict(int)
     with open(args.output_file, "w", encoding="utf-8") as f:

examples/preprocess/preprocess_para_pat.py

@@ -26,7 +26,7 @@ def get_args():
     )
     parser.add_argument(
         "--output_file",
-        default=(project_path / "data/para_pat.jsonl"),
+        default=(project_path / "data/para_pat_fr_ru.jsonl"),
         type=str
     )
 
@@ -38,12 +38,25 @@ def main():
     args = get_args()
 
     name_list = [
-        "cs-en", "de-en", "de-fr", "el-en", "en-es", "en-fr", "en-hu", "en-ja",
-        "en-ko", "en-pt", "en-ro", "en-ru", "en-sk",
-        "en-uk",
-        "en-zh",
-        "es-fr",
-        "fr-ja", "fr-ko", "fr-ru"
+        # "cs-en",
+        # "de-en",
+        # "de-fr",
+        # "el-en",
+        # "en-es",
+        # "en-fr",
+        # "en-hu",
+        # "en-ja",
+        # "en-ko",
+        # "en-pt",
+        # "en-ro",
+        # "en-ru",
+        # "en-sk",
+        # "en-uk",
+        # "en-zh",
+        # "es-fr",
+        # "fr-ja",
+        # "fr-ko",
+        "fr-ru"
     ]
 
     text_set = set()

language_identification.py

@@ -50,7 +50,27 @@ _URLS = {
     "para_crawl_en_pl": "data/para_crawl_en_pl.jsonl",
     "para_crawl_en_pt": "data/para_crawl_en_pt.jsonl",
     "para_crawl_en_ro": "data/para_crawl_en_ro.jsonl",
-    "para_pat": "data/para_pat.jsonl",
+    "para_crawl_en_sk": "data/para_crawl_en_sk.jsonl",
+    "para_crawl_en_sl": "data/para_crawl_en_sl.jsonl",
+    "para_crawl_en_sv": "data/para_crawl_en_sv.jsonl",
+    "para_pat_cs_en": "data/para_pat_cs_en.jsonl",
+    "para_pat_de_en": "data/para_pat_de_en.jsonl",
+    "para_pat_de_fr": "data/para_pat_de_fr.jsonl",
+    "para_pat_el_en": "data/para_pat_el_en.jsonl",
+    "para_pat_en_es": "data/para_pat_en_es.jsonl",
+    "para_pat_en_hu": "data/para_pat_en_hu.jsonl",
+    "para_pat_en_ja": "data/para_pat_en_ja.jsonl",
+    "para_pat_en_ko": "data/para_pat_en_ko.jsonl",
+    "para_pat_en_pt": "data/para_pat_en_pt.jsonl",
+    "para_pat_en_ro": "data/para_pat_en_ro.jsonl",
+    "para_pat_en_ru": "data/para_pat_en_ru.jsonl",
+    "para_pat_en_sk": "data/para_pat_en_sk.jsonl",
+    "para_pat_en_uk": "data/para_pat_en_uk.jsonl",
+    "para_pat_en_zh": "data/para_pat_en_zh.jsonl",
+    "para_pat_es_fr": "data/para_pat_es_fr.jsonl",
+    "para_pat_fr_ja": "data/para_pat_fr_ja.jsonl",
+    "para_pat_fr_ko": "data/para_pat_fr_ko.jsonl",
+    "para_pat_fr_ru": "data/para_pat_fr_ru.jsonl",
     "php": "data/php.jsonl",
     "scandi_langid": "data/scandi_langid.jsonl",
     "stsb_multi_mt": "data/stsb_multi_mt.jsonl",
@@ -177,7 +197,27 @@ class LanguageIdentification(datasets.GeneratorBasedBuilder):
         datasets.BuilderConfig(name="para_crawl_en_pl", version=VERSION, description="para_crawl_en_pl"),
         datasets.BuilderConfig(name="para_crawl_en_pt", version=VERSION, description="para_crawl_en_pt"),
         datasets.BuilderConfig(name="para_crawl_en_ro", version=VERSION, description="para_crawl_en_ro"),
-        datasets.BuilderConfig(name="para_pat", version=VERSION, description="para_pat"),
+        datasets.BuilderConfig(name="para_crawl_en_sk", version=VERSION, description="para_crawl_en_sk"),
+        datasets.BuilderConfig(name="para_crawl_en_sl", version=VERSION, description="para_crawl_en_sl"),
+        datasets.BuilderConfig(name="para_crawl_en_sv", version=VERSION, description="para_crawl_en_sv"),
+        datasets.BuilderConfig(name="para_pat_cs_en", version=VERSION, description="para_pat_cs_en"),
+        datasets.BuilderConfig(name="para_pat_de_en", version=VERSION, description="para_pat_de_en"),
+        datasets.BuilderConfig(name="para_pat_de_fr", version=VERSION, description="para_pat_de_fr"),
+        datasets.BuilderConfig(name="para_pat_el_en", version=VERSION, description="para_pat_el_en"),
+        datasets.BuilderConfig(name="para_pat_en_es", version=VERSION, description="para_pat_en_es"),
+        datasets.BuilderConfig(name="para_pat_en_hu", version=VERSION, description="para_pat_en_hu"),
+        datasets.BuilderConfig(name="para_pat_en_ja", version=VERSION, description="para_pat_en_ja"),
+        datasets.BuilderConfig(name="para_pat_en_ko", version=VERSION, description="para_pat_en_ko"),
+        datasets.BuilderConfig(name="para_pat_en_pt", version=VERSION, description="para_pat_en_pt"),
+        datasets.BuilderConfig(name="para_pat_en_ro", version=VERSION, description="para_pat_en_ro"),
+        datasets.BuilderConfig(name="para_pat_en_ru", version=VERSION, description="para_pat_en_ru"),
+        datasets.BuilderConfig(name="para_pat_en_sk", version=VERSION, description="para_pat_en_sk"),
+        datasets.BuilderConfig(name="para_pat_en_uk", version=VERSION, description="para_pat_en_uk"),
+        datasets.BuilderConfig(name="para_pat_en_zh", version=VERSION, description="para_pat_en_zh"),
+        datasets.BuilderConfig(name="para_pat_es_fr", version=VERSION, description="para_pat_es_fr"),
+        datasets.BuilderConfig(name="para_pat_fr_ja", version=VERSION, description="para_pat_fr_ja"),
+        datasets.BuilderConfig(name="para_pat_fr_ko", version=VERSION, description="para_pat_fr_ko"),
+        datasets.BuilderConfig(name="para_pat_fr_ru", version=VERSION, description="para_pat_fr_ru"),
         datasets.BuilderConfig(name="php", version=VERSION, description="php"),
         datasets.BuilderConfig(name="scandi_langid", version=VERSION, description="scandi_langid"),
         datasets.BuilderConfig(name="stsb_multi_mt", version=VERSION, description="stsb_multi_mt"),

load_data.md

@@ -111,22 +111,23 @@
 |   mt   |    maltese     |  100000   | multi_para_crawl |
 |   nl   |    dutch       |  100000   |       kde4       |
 |   no   |   norwegian    |  100000   | multi_para_crawl |
-|   pl   |    polish      |  -   |       ecb        |
-|   pt   |   portuguese   |  10000   |     tatoeba      |
-|   ro   |    romanian    |  10000   |       kde4       |
-|   ru   |    russian     |  10000   |       xnli       |
-|   sk   |    slovak      |  10000   | multi_para_crawl |
-|   sl   |   slovenian    |  4589    |   europa_ecdc_tm; europa_eac_tm   |
-|   sw   |    swahili     |  10000   |       xnli       |
-|   sv   |    swedish     |  10000   |       kde4       |
-|   th   |     thai       |  10000   |       xnli       |
-|   tl   |    tagalog     |  10000   | multi_para_crawl |
-|   tn   |    serpeti     |  10000   |   autshumato     |
-|   tr   |    turkish     |  10000   |       xnli       |
-|   ts   |    dzonga      |  10000   |    autshumato    |
-|   ur   |     urdu       |  10000   |       xnli       |
-|   vi   |   vietnamese   |  10000   |       xnli       |
+|   pl   |    polish      |  100000   | para_crawl_en_pl |
+|   pt   |   portuguese   |  100000   | para_crawl_en_pt |
+|   ro   |    romanian    |  100000   |     iwslt2017    |
+|   ru   |    russian     |  100000   |       xnli       |
+|   sk   |    slovak      |  100000   | multi_para_crawl |
+|   sl   |   slovenian    |  100000   | para_crawl_en_sl |
+|   sw   |    swahili     |  100000   |       xnli       |
+|   sv   |    swedish     |  100000   |       kde4       |
+|   th   |     thai       |  100000   |       xnli       |
+|   tl   |    tagalog     |  97241    | multi_para_crawl |
+|   tn   |    serpeti     |  100000   |   autshumato     |
+|   tr   |    turkish     |  100000   |       xnli       |
+|   ts   |    dzonga      |  100000   |    autshumato    |
+|   uk   |    ukrainian   |  88533    |  para_pat_en_uk  |
+|   ur   |     urdu       |  100000   |       xnli       |
+|   vi   |   vietnamese   |  100000   |       xnli       |
 |   yo   |     yoruba     |  9970    |    menyo20k_mt   |
-|   zh   |    chinese     |  10000   |       xnli       |
-|   zu   |  zulu, south africa  |  10000   |    autshumato    |
+|   zh   |    chinese     |  100000   |       xnli       |
+|   zu   |  zulu, south africa  |  26801   |    autshumato    |