Card: record the toolkit and model behind each language

#3
by jknafou - opened
Files changed (1) hide show
  1. README.md +33 -8
README.md CHANGED
@@ -31,6 +31,31 @@ tags:
31
  pretty_name: TransCorpus-bio
32
  size_categories:
33
  - 100M<n<1B
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
34
  configs:
35
  - config_name: eng_Latn
36
  default: true
@@ -64,19 +89,19 @@ configs:
64
 
65
  # TransCorpus-bio
66
 
67
- **TransCorpus-bio** is a large-scale, parallel biomedical corpus consisting of PubMed abstracts (title + abstract), translated with the [TransCorpus Toolkit](https://github.com/jknafou/TransCorpus) using NLLB-200. It is designed to enable high-quality multi-lingual biomedical language modeling and downstream NLP research.
68
 
69
  This dataset was restructured from five separate single-language repositories into one dataset with a config (tab in the dataset viewer) per language, and with each row carrying its source PMID.
70
 
71
  ## Languages
72
 
73
- | Config | Language | Rows |
74
- |---|---|---|
75
- | `eng_Latn` (default) | English | 21,567,136 |
76
- | `spa_Latn` | Spanish | 21,567,136 |
77
- | `fra_Latn` | French | 21,567,136 |
78
- | `deu_Latn` | German | 21,567,136 |
79
- | `hin_Deva` | Hindi | 21,567,136 |
80
 
81
  Every config has the same row count and the same schema: `pmid` (int64) and `text` (string, title + abstract on one line). The `pmid` column is the original PubMed ID of the source English abstract, is identical across every language for the same document, and is unique within each config, so joining any two configs on `pmid` reconstructs an aligned parallel pair. Row order and row count were verified to match exactly across all five languages and the source PMID list before publishing.
82
 
 
31
  pretty_name: TransCorpus-bio
32
  size_categories:
33
  - 100M<n<1B
34
+ translations:
35
+ - config_name: spa_Latn
36
+ source_config: eng_Latn
37
+ toolkit: TransCorpus Toolkit
38
+ toolkit_url: https://github.com/jknafou/TransCorpus
39
+ model: M2M-100 1.2B
40
+ model_url: https://huggingface.co/facebook/m2m100_1.2B
41
+ - config_name: fra_Latn
42
+ source_config: eng_Latn
43
+ toolkit: TransCorpus Toolkit
44
+ toolkit_url: https://github.com/jknafou/TransCorpus
45
+ model: M2M-100 1.2B
46
+ model_url: https://huggingface.co/facebook/m2m100_1.2B
47
+ - config_name: deu_Latn
48
+ source_config: eng_Latn
49
+ toolkit: TransCorpus Toolkit
50
+ toolkit_url: https://github.com/jknafou/TransCorpus
51
+ model: M2M-100 1.2B
52
+ model_url: https://huggingface.co/facebook/m2m100_1.2B
53
+ - config_name: hin_Deva
54
+ source_config: eng_Latn
55
+ toolkit: TransCorpus Toolkit
56
+ toolkit_url: https://github.com/jknafou/TransCorpus
57
+ model: M2M-100 1.2B
58
+ model_url: https://huggingface.co/facebook/m2m100_1.2B
59
  configs:
60
  - config_name: eng_Latn
61
  default: true
 
89
 
90
  # TransCorpus-bio
91
 
92
+ **TransCorpus-bio** is a large-scale, parallel biomedical corpus consisting of PubMed abstracts (title + abstract), machine-translated from English. Each language records the toolkit and the model that produced it (the table below, and the `translations` field of this card's metadata). It is designed to enable high-quality multi-lingual biomedical language modeling and downstream NLP research.
93
 
94
  This dataset was restructured from five separate single-language repositories into one dataset with a config (tab in the dataset viewer) per language, and with each row carrying its source PMID.
95
 
96
  ## Languages
97
 
98
+ | Config | Language | Rows | Translated with | Model |
99
+ |---|---|---|---|---|
100
+ | `eng_Latn` (default) | English | 21,567,136 | source (original PubMed text) | – |
101
+ | `spa_Latn` | Spanish | 21,567,136 | [TransCorpus Toolkit](https://github.com/jknafou/TransCorpus) | [M2M-100 1.2B](https://huggingface.co/facebook/m2m100_1.2B) |
102
+ | `fra_Latn` | French | 21,567,136 | [TransCorpus Toolkit](https://github.com/jknafou/TransCorpus) | [M2M-100 1.2B](https://huggingface.co/facebook/m2m100_1.2B) |
103
+ | `deu_Latn` | German | 21,567,136 | [TransCorpus Toolkit](https://github.com/jknafou/TransCorpus) | [M2M-100 1.2B](https://huggingface.co/facebook/m2m100_1.2B) |
104
+ | `hin_Deva` | Hindi | 21,567,136 | [TransCorpus Toolkit](https://github.com/jknafou/TransCorpus) | [M2M-100 1.2B](https://huggingface.co/facebook/m2m100_1.2B) |
105
 
106
  Every config has the same row count and the same schema: `pmid` (int64) and `text` (string, title + abstract on one line). The `pmid` column is the original PubMed ID of the source English abstract, is identical across every language for the same document, and is unique within each config, so joining any two configs on `pmid` reconstructs an aligned parallel pair. Row order and row count were verified to match exactly across all five languages and the source PMID list before publishing.
107