Dataset Card for No Language Left Behind (NLLB - 200vo)

Dataset Summary

This dataset was created based on metadata for mined bitext released by Meta AI. It contains bitext for 148 English-centric and 1465 non-English-centric language pairs using the stopes mining library and the LASER3 encoders (Heffernan et al., 2022). The complete dataset is ~450GB.

CCMatrix contains previous versions of mined instructions.

How to use the data

There are two ways to access the data:

• Via the Hugging Face Python datasets library

For accessing a particular language pair:

from datasets import load_dataset
dataset = load_dataset("allenai/nllb", "ace_Latn-ban_Latn")

• Clone the git repo

git lfs install
git clone https://huggingface.co/datasets/allenai/nllb

Supported Tasks and Leaderboards

N/A

Languages

Language pairs can be found here.

Dataset Structure

The dataset contains gzipped tab delimited text files for each direction. Each text file contains lines with parallel sentences.

Data Instances

The number of instances for each language pair can be found in the dataset_infos.json file.

Data Fields

Every instance for a language pair contains the following fields: 'translation' (containing sentence pairs), 'laser_score', 'source_sentence_lid', 'target_sentence_lid', where 'lid' is language classification probability, 'source_sentence_source', 'source_sentence_url', 'target_sentence_source', 'target_sentence_url'.

• Sentence in first language
• Sentence in second language
• LASER score
• Language ID score for first sentence
• Language ID score for second sentence
• First sentence source (See Source Data Table)
• First sentence URL if the source is crawl-data/*; _ otherwise
• Second sentence source
• Second sentence URL if the source is crawl-data/*; _ otherwise

The lines are sorted by LASER3 score in decreasing order.

Example:

{'translation': {'ace_Latn': 'Gobnyan hana geupeukeucewa gata atawa geutinggai meunan mantong gata."',
  'ban_Latn': 'Ida nenten jaga manggayang wiadin ngutang semeton."'},
 'laser_score': 1.2499876022338867,
 'source_sentence_lid': 1.0000100135803223,
 'target_sentence_lid': 0.9991400241851807,
 'source_sentence_source': 'paracrawl9_hieu',
 'source_sentence_url': '_',
 'target_sentence_source': 'crawl-data/CC-MAIN-2020-10/segments/1581875144165.4/wet/CC-MAIN-20200219153707-20200219183707-00232.warc.wet.gz',
 'target_sentence_url': 'https://alkitab.mobi/tb/Ula/31/6/\n'}

Data Splits

The data is not split. Given the noisy nature of the overall process, we recommend using the data only for training and use other datasets like Flores-200 for the evaluation. The data includes some development and test sets from other datasets, such as xlsum. In addition, sourcing data from multiple web crawls is likely to produce incidental overlap with other test sets.

Dataset Creation

Curation Rationale

Data was filtered based on language identification, emoji based filtering, and for some high-resource languages using a language model. For more details on data filtering please refer to Section 5.2 (NLLB Team et al., 2022).

Source Data

Initial Data Collection and Normalization

Monolingual data was collected from the following sources:

Name in data	Source
afriberta	https://github.com/castorini/afriberta
americasnlp	https://github.com/AmericasNLP/americasnlp2021/
bho_resources	https://github.com/shashwatup9k/bho-resources
crawl-data/*	WET files from https://commoncrawl.org/the-data/get-started/
emcorpus	http://lepage-lab.ips.waseda.ac.jp/en/projects/meiteilon-manipuri-language-resources/
fbseed20220317	https://github.com/facebookresearch/flores/tree/main/nllb_seed
giossa_mono	https://github.com/sgongora27/giossa-gongora-guarani-2021
iitguwahati	https://github.com/priyanshu2103/Sanskrit-Hindi-Machine-Translation/tree/main/parallel-corpus
indic	https://indicnlp.ai4bharat.org/corpora/
lacunaner	https://github.com/masakhane-io/lacuna_pos_ner/tree/main/language_corpus
leipzig	Community corpora from https://wortschatz.uni-leipzig.de/en/download for each year available
lowresmt2020	https://github.com/panlingua/loresmt-2020
masakhanener	https://github.com/masakhane-io/masakhane-ner/tree/main/MasakhaNER2.0/data
nchlt	https://repo.sadilar.org/handle/20.500.12185/299 https://repo.sadilar.org/handle/20.500.12185/302 https://repo.sadilar.org/handle/20.500.12185/306 https://repo.sadilar.org/handle/20.500.12185/308 https://repo.sadilar.org/handle/20.500.12185/309 https://repo.sadilar.org/handle/20.500.12185/312 https://repo.sadilar.org/handle/20.500.12185/314 https://repo.sadilar.org/handle/20.500.12185/315 https://repo.sadilar.org/handle/20.500.12185/321 https://repo.sadilar.org/handle/20.500.12185/325 https://repo.sadilar.org/handle/20.500.12185/328 https://repo.sadilar.org/handle/20.500.12185/330 https://repo.sadilar.org/handle/20.500.12185/332 https://repo.sadilar.org/handle/20.500.12185/334 https://repo.sadilar.org/handle/20.500.12185/336 https://repo.sadilar.org/handle/20.500.12185/337 https://repo.sadilar.org/handle/20.500.12185/341 https://repo.sadilar.org/handle/20.500.12185/343 https://repo.sadilar.org/handle/20.500.12185/346 https://repo.sadilar.org/handle/20.500.12185/348 https://repo.sadilar.org/handle/20.500.12185/353 https://repo.sadilar.org/handle/20.500.12185/355 https://repo.sadilar.org/handle/20.500.12185/357 https://repo.sadilar.org/handle/20.500.12185/359 https://repo.sadilar.org/handle/20.500.12185/362 https://repo.sadilar.org/handle/20.500.12185/364
paracrawl-2022-*	https://data.statmt.org/paracrawl/monolingual/
paracrawl9*	https://paracrawl.eu/moredata the monolingual release
pmi	https://data.statmt.org/pmindia/
til	https://github.com/turkic-interlingua/til-mt/tree/master/til_corpus
w2c	https://lindat.mff.cuni.cz/repository/xmlui/handle/11858/00-097C-0000-0022-6133-9
xlsum	https://github.com/csebuetnlp/xl-sum

Who are the source language producers?

Text was collected from the web and various monolingual data sets, many of which are also web crawls. This may have been written by people, generated by templates, or in some cases be machine translation output.

Annotations

Annotation process

Parallel sentences in the monolingual data were identified using LASER3 encoders. (Heffernan et al., 2022)

Who are the annotators?

The data was not human annotated.

Personal and Sensitive Information

Data may contain personally identifiable information, sensitive content, or toxic content that was publicly shared on the Internet.

Considerations for Using the Data

Social Impact of Dataset

This dataset provides data for training machine learning systems for many languages that have low resources available for NLP.

Discussion of Biases

Biases in the data have not been specifically studied, however as the original source of data is World Wide Web it is likely that the data has biases similar to those prevalent in the Internet. The data may also exhibit biases introduced by language identification and data filtering techniques; lower resource languages generally have lower accuracy.

Other Known Limitations

Some of the translations are in fact machine translations. While some website machine translation tools are identifiable from HTML source, these tools were not filtered out en mass because raw HTML was not available from some sources and CommonCrawl processing started from WET files.

Additional Information

Dataset Curators

The data was not curated.

Licensing Information

The dataset is released under the terms of ODC-BY. By using this, you are also bound to the respective Terms of Use and License of the original source.

Citation Information

Schwenk et al, CCMatrix: Mining Billions of High-Quality Parallel Sentences on the Web. ACL https://aclanthology.org/2021.acl-long.507/ Hefferman et al, Bitext Mining Using Distilled Sentence Representations for Low-Resource Languages. Arxiv https://arxiv.org/abs/2205.12654, 2022.
NLLB Team et al, No Language Left Behind: Scaling Human-Centered Machine Translation, Arxiv https://arxiv.org/abs/2207.04672, 2022.

Contributions

We thank the NLLB Meta AI team for open sourcing the meta data and instructions on how to use it with special thanks to Bapi Akula, Pierre Andrews, Onur Çelebi, Sergey Edunov, Kenneth Heafield, Philipp Koehn, Alex Mourachko, Safiyyah Saleem, Holger Schwenk, and Guillaume Wenzek. We also thank the AllenNLP team at AI2 for hosting and releasing this data, including Akshita Bhagia (for engineering efforts to host the data, and create the huggingface dataset), and Jesse Dodge (for organizing the connection).