Scaling neural machine translation to 200 languages
Authors
January 1, 2024
Abstract
Presents the No Language Left Behind multilingual machine translation system. The released model family explicitly supports Central Kurdish in Arabic script and Northern Kurdish in Latin script among approximately 200 supported languages.
Keywords
Neural Machine Translation
NLLB
Central Kurdish
Northern Kurdish
Multilingual Translation
Related Datasets
FLORES-200 Kurdish Evaluation Benchmark
2022
3,001 designed sentences per language; 2,009 publicly released dev/devtest sentences per Kurdish variety
TXT
Multilingual machine translation evaluation benchmark containing Central Kurdish (ckb_Arab) and Northern Kurdish (kmr_Latn). Public Kurdish portions contain 997 development and 1,012 development-test …
NLLB-200 Translation Model Family
2022
Representative release: approximately 600M parameters
PyTorch, SentencePiece, JSON
Released multilingual neural machine translation model family supporting Central Kurdish (ckb_Arab) and Northern Kurdish (kmr_Latn). The 600M distilled checkpoint is a representative publicly …
Citation
NLLB Team. (2024). Scaling neural machine translation to 200 languages. Nature, 630, 841-846.