Hugging Face's logo

Datasets:
openlanguagedata
/
oldi_seed

Tasks:
Text Generation
Translation
Modalities:
Text
Formats:
parquet
Languages:
Achinese
Moroccan Arabic
Egyptian Arabic
Size:
100K - 1M
ArXiv:
Libraries:
Datasets
Dask
License:
Dataset card Data Studio Files
xet
Community
5
laurievb commited on
Commit
8723273
·
verified ·
1 Parent(s): bd6e2eb

Alter to match FLORES plus readme

Browse files
Files changed (1) hide show
  1. README.md +149 -99
README.md CHANGED
@@ -1,99 +1,149 @@
1
- ---
2
- annotations_creators:
3
- - found
4
- language_creators:
5
- - expert-generated
6
- license: cc-by-sa-4.0
7
- pretty_name: OLDI-Seed
8
- task_categories:
9
- - text2text-generation
10
- - translation
11
- configs:
12
- - config_name: default
13
- data_files:
14
- - split: train
15
- path: seed/*.parquet
16
- ---
17
- # OLDI Seed Machine Translation Data
18
-
19
- This repository hosts the open source seed dataset described in the paper [Small Data, Big Impact: Leveraging Minimal Data for Effective Machine Translation](https://aclanthology.org/2023.acl-long.154). The data, which is licensed under [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/), is currently being managed by OLDI, the [Open Language Data Initiative](https://oldi.org).
20
-
21
- ## Data
22
-
23
- The parallel corpus currently contains 6,193 sentences sampled from English Wikipedia and translated into 44 languages. It can be used to kick-start machine translation models for language directions which currently lack large-scale datasets.
24
-
25
- ## Contributing
26
-
27
- Fixes and new language contributions are most welcome. Please see the [Contribution guidelines](https://oldi.org/guidelines) for further information.
28
-
29
- ## Language coverage
30
-
31
- | Code | Script | Glottocode | Name | Notes |
32
- |-------|--------|------------|-----------------------------------|------------------------------------------------------------|
33
- | `ace` | `Arab` | `achi1257` | Acehnese (Jawi script) | |
34
- | `ace` | `Latn` | `achi1257` | Acehnese (Latin script) | |
35
- | `ary` | `Arab` | `moro1292` | Moroccan Arabic | |
36
- | `arz` | `Arab` | `egyp1253` | Egyptian Arabic | |
37
- | `bam` | `Latn` | `bamb1269` | Bambara | |
38
- | `ban` | `Latn` | `bali1278` | Balinese | |
39
- | `ben` | `Beng` | `beng1280` | Bengali | |
40
- | `bho` | `Deva` | `bhoj1244` | Bhojpuri | |
41
- | `bjn` | `Arab` | `banj1239` | Banjar (Jawi script) | |
42
- | `bjn` | `Latn` | `banj1239` | Banjar (Latin script) | |
43
- | `bug` | `Latn` | `bugi1244` | Buginese | |
44
- | `crh` | `Latn` | `crim1257` | Crimean Tatar | |
45
- | `dik` | `Latn` | `sout2832` | Southwestern Dinka | |
46
- | `dzo` | `Tibt` | `dzon1239` | Dzongkha | |
47
- | `eng` | `Latn` | `stan1293` | English | |
48
- | `fur` | `Latn` | `east2271` | Friulian | |
49
- | `fuv` | `Latn` | `nige1253` | Nigerian Fulfulde | |
50
- | `gug` | `Latn` | `para1311` | Paraguayan Guaraní | |
51
- | `hne` | `Deva` | `chha1249` | Chhattisgarhi | |
52
- | `ita` | `Latn` | `ital1282` | Italian | |
53
- | `kas` | `Arab` | `kash1277` | Kashmiri (Arabic script) | |
54
- | `kas` | `Deva` | `kash1277` | Kashmiri (Devanagari script) | |
55
- | `knc` | `Arab` | `cent2050` | Central Kanuri (Arabic script) | |
56
- | `knc` | `Latn` | `cent2050` | Central Kanuri (Latin script) | |
57
- | `lij` | `Latn` | `geno1240` | Ligurian (Genoese) | |
58
- | `lim` | `Latn` | `limb1263` | Limburgish | |
59
- | `lmo` | `Latn` | `lomb1257` | Lombard | [[1]](https://github.com/openlanguagedata/flores/issues/5) |
60
- | `ltg` | `Latn` | `east2282` | Latgalian | |
61
- | `mag` | `Deva` | `maga1260` | Magahi | |
62
- | `mni` | `Beng` | `mani1292` | Meitei (Manipuri, Bengali script) | |
63
- | `mri` | `Latn` | `maor1246` | Maori | |
64
- | `nqo` | `Nkoo` | `nkoa1234` | Nko | |
65
- | `nus` | `Latn` | `nuer1246` | Nuer | |
66
- | `pbt` | `Arab` | `sout2649` | Southern Pashto | |
67
- | `prs` | `Arab` | `dari1249` | Dari | |
68
- | `scn` | `Latn` | `sici1248` | Sicilian | |
69
- | `shn` | `Mymr` | `shan1277` | Shan | |
70
- | `spa` | `Latn` | `amer1254` | Spanish (Latin American) | |
71
- | `srd` | `Latn` | `sard1257` | Sardinian | [[1]](https://github.com/openlanguagedata/flores/issues/6) |
72
- | `szl` | `Latn` | `sile1253` | Silesian | |
73
- | `taq` | `Latn` | `tama1365` | Tamasheq (Latin script) | |
74
- | `taq` | `Tfng` | `tama1365` | Tamasheq (Tifinagh script) | |
75
- | `vec` | `Latn` | `vene1259` | Venetian | |
76
- | `zgh` | `Tfng` | `stan1324` | Standard Moroccan Tamazight | |
77
-
78
- ## Citation
79
-
80
- This dataset is based upon the work originally described in the following paper:
81
-
82
- ```bibtex
83
- @inproceedings{seed-23,
84
- title = {Small Data, Big Impact: Leveraging Minimal Data for Effective Machine Translation},
85
- author = {Maillard, Jean and Gao, Cynthia and Kalbassi, Elahe and Sadagopan, Kaushik Ram and Goswami, Vedanuj and Koehn, Philipp and Fan, Angela and Guzmán, Francisco},
86
- booktitle = {Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
87
- year = {2023},
88
- address = {Toronto, Canada},
89
- publisher = {Association for Computational Linguistics},
90
- pages = {2740--2756},
91
- url = {https://aclanthology.org/2023.acl-long.154},
92
- }
93
- ```
94
-
95
- Other authors have since contributed to the dataset. If you use this dataset in your work, please cite the relevant papers listed in [bibliography.bib](bibliography.bib).
96
-
97
- ## Changelog
98
-
99
- See [CHANGELOG.md](CHANGELOG.md) for information about the latest changes.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ annotations_creators:
3
+ - found
4
+ language_creators:
5
+ - expert-generated
6
+ license: cc-by-sa-4.0
7
+ pretty_name: OLDI-Seed
8
+ task_categories:
9
+ - text2text-generation
10
+ - translation
11
+ configs:
12
+ - config_name: default
13
+ data_files:
14
+ - split: train
15
+ path: seed/*.parquet
16
+ ---
17
+ # OLDI Seed Machine Translation Datacard
18
+
19
+ OLDI Seed is a machine translation dataset designed to be used to kick-start machine translation models for language directions which currently lack large-scale datasets.
20
+
21
+
22
+ ## Dataset Details
23
+
24
+ ### Dataset Description
25
+
26
+ OLDI Seed is a parallel corpus which consists of 6,193 sentences sampled from English Wikipedia and translated into 44 languages. It can be used to kick-start machine translation models for language directions which currently lack large-scale datasets.
27
+
28
+ The data, which is licensed under [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/), is currently being managed by OLDI, the [Open Language Data Initiative](https://oldi.org).
29
+
30
+ - **Curated by:** [The Open Language Data Initiative](https://oldi.org/)
31
+ - **Language(s) (NLP):** Around 40, see documentation for full list.
32
+ - **License:** [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/)
33
+
34
+ ### Dataset Sources
35
+
36
+ OLDI Seed is an updated version of the open source seed dataset described in the paper [Small Data, Big Impact: Leveraging Minimal Data for Effective Machine Translation](https://aclanthology.org/2023.acl-long.154):
37
+
38
+ ```bibtex
39
+ @inproceedings{seed-23,
40
+ title = {Small Data, Big Impact: Leveraging Minimal Data for Effective Machine Translation},
41
+ author = {Maillard, Jean and Gao, Cynthia and Kalbassi, Elahe and Sadagopan, Kaushik Ram and Goswami, Vedanuj and Koehn, Philipp and Fan, Angela and Guzmán, Francisco},
42
+ booktitle = {Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
43
+ year = {2023},
44
+ address = {Toronto, Canada},
45
+ publisher = {Association for Computational Linguistics},
46
+ pages = {2740--2756},
47
+ url = {https://aclanthology.org/2023.acl-long.154},
48
+ }
49
+ ```
50
+
51
+ Other authors have since contributed to the dataset. If you use this dataset in your work, please cite the relevant papers listed in [bibliography.bib](bibliography.bib).
52
+
53
+ ### Language coverage
54
+ | Code | Script | Glottocode | Name | Notes |
55
+ |-------|--------|------------|-----------------------------------|------------------------------------------------------------|
56
+ | `ace` | `Arab` | `achi1257` | Acehnese (Jawi script) | |
57
+ | `ace` | `Latn` | `achi1257` | Acehnese (Latin script) | |
58
+ | `ary` | `Arab` | `moro1292` | Moroccan Arabic | |
59
+ | `arz` | `Arab` | `egyp1253` | Egyptian Arabic | |
60
+ | `bam` | `Latn` | `bamb1269` | Bambara | |
61
+ | `ban` | `Latn` | `bali1278` | Balinese | |
62
+ | `ben` | `Beng` | `beng1280` | Bengali | |
63
+ | `bho` | `Deva` | `bhoj1244` | Bhojpuri | |
64
+ | `bjn` | `Arab` | `banj1239` | Banjar (Jawi script) | |
65
+ | `bjn` | `Latn` | `banj1239` | Banjar (Latin script) | |
66
+ | `bug` | `Latn` | `bugi1244` | Buginese | |
67
+ | `crh` | `Latn` | `crim1257` | Crimean Tatar | |
68
+ | `dik` | `Latn` | `sout2832` | Southwestern Dinka | |
69
+ | `dzo` | `Tibt` | `dzon1239` | Dzongkha | |
70
+ | `eng` | `Latn` | `stan1293` | English | |
71
+ | `fur` | `Latn` | `east2271` | Friulian | |
72
+ | `fuv` | `Latn` | `nige1253` | Nigerian Fulfulde | |
73
+ | `gug` | `Latn` | `para1311` | Paraguayan Guaraní | |
74
+ | `hne` | `Deva` | `chha1249` | Chhattisgarhi | |
75
+ | `ita` | `Latn` | `ital1282` | Italian | |
76
+ | `kas` | `Arab` | `kash1277` | Kashmiri (Arabic script) | |
77
+ | `kas` | `Deva` | `kash1277` | Kashmiri (Devanagari script) | |
78
+ | `knc` | `Arab` | `cent2050` | Central Kanuri (Arabic script) | |
79
+ | `knc` | `Latn` | `cent2050` | Central Kanuri (Latin script) | |
80
+ | `lij` | `Latn` | `geno1240` | Ligurian (Genoese) | |
81
+ | `lim` | `Latn` | `limb1263` | Limburgish | |
82
+ | `lmo` | `Latn` | `lomb1257` | Lombard | [[1]](https://github.com/openlanguagedata/flores/issues/5) |
83
+ | `ltg` | `Latn` | `east2282` | Latgalian | |
84
+ | `mag` | `Deva` | `maga1260` | Magahi | |
85
+ | `mni` | `Beng` | `mani1292` | Meitei (Manipuri, Bengali script) | |
86
+ | `mri` | `Latn` | `maor1246` | Maori | |
87
+ | `nqo` | `Nkoo` | `nkoa1234` | Nko | |
88
+ | `nus` | `Latn` | `nuer1246` | Nuer | |
89
+ | `pbt` | `Arab` | `sout2649` | Southern Pashto | |
90
+ | `prs` | `Arab` | `dari1249` | Dari | |
91
+ | `scn` | `Latn` | `sici1248` | Sicilian | |
92
+ | `shn` | `Mymr` | `shan1277` | Shan | |
93
+ | `spa` | `Latn` | `amer1254` | Spanish (Latin American) | |
94
+ | `srd` | `Latn` | `sard1257` | Sardinian | [[1]](https://github.com/openlanguagedata/flores/issues/6) |
95
+ | `szl` | `Latn` | `sile1253` | Silesian | |
96
+ | `taq` | `Latn` | `tama1365` | Tamasheq (Latin script) | |
97
+ | `taq` | `Tfng` | `tama1365` | Tamasheq (Tifinagh script) | |
98
+ | `vec` | `Latn` | `vene1259` | Venetian | |
99
+ | `zgh` | `Tfng` | `stan1324` | Standard Moroccan Tamazight | |
100
+
101
+ ## Uses
102
+
103
+ OLDI Seed is intended to be used to kick-start machine translation models for language directions which currently lack large-scale datasets.
104
+
105
+ ## Dataset Structure
106
+
107
+ Each instance in the dataset is structured as in the following example:
108
+
109
+ ```json
110
+ {
111
+ "id": 17,
112
+ "iso_639_3": "gug",
113
+ "iso_15924": "Latn",
114
+ "glottocode": "para1311",
115
+ "text": "Osẽ emperatriz viuda Maria Feodorovna-ramo peteĩ musical de Broadway mbykymi héravaa Anya 1965 arýpe.",
116
+ "url": "https://en.wikipedia.org/wiki/Lillian_Gish",
117
+ "last_updated": "1.0"
118
+ }
119
+ ```
120
+
121
+ ### Data Fields
122
+ - `id`: ID number for each line of data. Lines with the same ID in the same split are translations of each other.
123
+ - `iso_639_3`: The ISO 639-3 code indicating the language variety.
124
+ - `iso_15924`: The ISO 15924 code indicating the script.
125
+ - `glottocode`: The [Glottocode](https://glottolog.org/glottolog/language) corresponding to the language variety.
126
+ - `text`: A line of text in the indicated language.
127
+ - `url`: The URL for the English article from which the sentence was extracted.
128
+ - `last_updated`: The version number of the dataset in which the entry was last updated.
129
+
130
+ ## Dataset Creation
131
+
132
+ See the paper [Small Data, Big Impact: Leveraging Minimal Data for Effective Machine Translation](https://aclanthology.org/2023.acl-long.154) and the [NLLB paper](https://arxiv.org/pdf/2207.04672#page=18.71) for more details.
133
+
134
+ ### Additional Datacards
135
+
136
+ The datasets for some language varieties have individual datacards describing their creation. These can be found in the [dataset_cards](https://huggingface.co/datasets/openlanguagedata/oldi_seed/tree/main/dataset_cards) directory.
137
+
138
+ ## Dataset Card Contact
139
+
140
+ For more information about the OLDI Seed dataset, please see [oldi.org](https://oldi.org/)
141
+
142
+ ## Contributing
143
+
144
+ Fixes and new language contributions are most welcome. Please see the [Contribution guidelines](https://oldi.org/guidelines) for further information.
145
+
146
+ ## Changelog
147
+
148
+ See [CHANGELOG.md](CHANGELOG.md) for information about the latest changes.
149
+