metadata

annotations_creators:
  - found
language_creators:
  - expert-generated
license: cc-by-sa-4.0
pretty_name: OLDI-Seed
task_categories:
  - text2text-generation
  - translation
configs:
  - config_name: default
    data_files:
      - split: train
        path: seed/*.parquet

OLDI Seed Machine Translation Datacard

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.

Dataset Details

Dataset Description

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.

The data, which is licensed under CC BY-SA 4.0, is currently being managed by OLDI, the Open Language Data Initiative.

Curated by: The Open Language Data Initiative
Language(s) (NLP): Around 40, see documentation for full list.
License: CC BY-SA 4.0

Dataset Sources

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:

@inproceedings{seed-23,
    title = {Small Data, Big Impact: Leveraging Minimal Data for Effective Machine Translation},
    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},
    booktitle = {Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
    year = {2023},
    address = {Toronto, Canada},
    publisher = {Association for Computational Linguistics},
    pages = {2740--2756},
    url = {https://aclanthology.org/2023.acl-long.154},
}

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.

Uses

OLDI Seed is intended to be used to kick-start machine translation models for language directions which currently lack large-scale datasets.

Dataset Structure

Each instance in the dataset is structured as in the following example:

{
  "id": 17,
  "iso_639_3": "gug",
  "iso_15924": "Latn",
  "glottocode": "para1311",
  "text": "Osẽ emperatriz viuda Maria Feodorovna-ramo peteĩ musical de Broadway mbykymi héravaa Anya 1965 arýpe.",
  "url": "https://en.wikipedia.org/wiki/Lillian_Gish",
  "last_updated": "1.0"
}

Data Fields

id: ID number for each line of data. Lines with the same ID in the same split are translations of each other.
iso_639_3: The ISO 639-3 code indicating the language variety.
iso_15924: The ISO 15924 code indicating the script.
glottocode: The Glottocode corresponding to the language variety.
text: A line of text in the indicated language.
url: The URL for the English article from which the sentence was extracted.
last_updated: The version number of the dataset in which the entry was last updated.

Dataset Creation

See the paper Small Data, Big Impact: Leveraging Minimal Data for Effective Machine Translation and the NLLB paper for more details.

Additional Datacards

The datasets for some language varieties have individual datacards describing their creation. These can be found in the dataset_cards directory.

Dataset Card Contact

For more information about the OLDI Seed dataset, please see oldi.org

Contributing

Fixes and new language contributions are most welcome.

By contributing to this project you agree to the Developer Certificate of Origin (DCO). This document was created by the Linux Kernel community and is a simple statement that you, as a contributor, have the legal right to make the contribution.

In order to show your agreement with the DCO you should include at the end of commit message, the following line: Signed-off-by: John Doe <john.doe@example.com>, using your real name.

This can be done easily using the -s flag on the git commit.

Please see the Contribution guidelines for further information.

Changelog

See CHANGELOG.md for information about the latest changes.

Language coverage

Code	Script	Glottocode	Name	Notes
`ace`	`Arab`	`achi1257`	Acehnese (Jawi script)
`ace`	`Latn`	`achi1257`	Acehnese (Latin script)
`ary`	`Arab`	`moro1292`	Moroccan Arabic
`arz`	`Arab`	`egyp1253`	Egyptian Arabic
`bam`	`Latn`	`bamb1269`	Bambara
`ban`	`Latn`	`bali1278`	Balinese
`ben`	`Beng`	`beng1280`	Bengali
`bho`	`Deva`	`bhoj1244`	Bhojpuri
`bjn`	`Arab`	`banj1239`	Banjar (Jawi script)
`bjn`	`Latn`	`banj1239`	Banjar (Latin script)
`bug`	`Latn`	`bugi1244`	Buginese
`crh`	`Latn`	`crim1257`	Crimean Tatar
`dik`	`Latn`	`sout2832`	Southwestern Dinka
`dzo`	`Tibt`	`dzon1239`	Dzongkha
`eng`	`Latn`	`stan1293`	English
`fur`	`Latn`	`east2271`	Friulian
`fuv`	`Latn`	`nige1253`	Nigerian Fulfulde
`gug`	`Latn`	`para1311`	Paraguayan Guaraní
`hne`	`Deva`	`chha1249`	Chhattisgarhi
`ita`	`Latn`	`ital1282`	Italian
`kas`	`Arab`	`kash1277`	Kashmiri (Arabic script)
`kas`	`Deva`	`kash1277`	Kashmiri (Devanagari script)
`knc`	`Arab`	`cent2050`	Central Kanuri (Arabic script)
`knc`	`Latn`	`cent2050`	Central Kanuri (Latin script)
`lij`	`Latn`	`geno1240`	Ligurian (Genoese)
`lim`	`Latn`	`limb1263`	Limburgish
`lmo`	`Latn`	`lomb1257`	Lombard	[1]
`ltg`	`Latn`	`east2282`	Latgalian
`mag`	`Deva`	`maga1260`	Magahi
`mni`	`Beng`	`mani1292`	Meitei (Manipuri, Bengali script)
`mri`	`Latn`	`maor1246`	Maori
`nqo`	`Nkoo`	`nkoa1234`	Nko
`nus`	`Latn`	`nuer1246`	Nuer
`pbt`	`Arab`	`sout2649`	Southern Pashto
`prs`	`Arab`	`dari1249`	Dari
`scn`	`Latn`	`sici1248`	Sicilian
`shn`	`Mymr`	`shan1277`	Shan
`spa`	`Latn`	`amer1254`	Spanish (Latin American)
`srd`	`Latn`	`sard1257`	Sardinian	[1]
`szl`	`Latn`	`sile1253`	Silesian
`taq`	`Latn`	`tama1365`	Tamasheq (Latin script)
`taq`	`Tfng`	`tama1365`	Tamasheq (Tifinagh script)
`vec`	`Latn`	`vene1259`	Venetian
`zgh`	`Tfng`	`stan1324`	Standard Moroccan Tamazight