We introduce our efforts towards building a universal neural machine translation (NMT) system capable of translating between any language pair. We set a milestone towards this goal by building a single massively multilingual NMT model handling 103 languages trained on over 25 billion examples. Our system demonstrates effective transfer learning ability, significantly improving translation quality of low-resource languages, while keeping high-resource language translation quality on-par with competitive bilingual baselines. We provide in-depth analysis of various aspects of model building that are crucial to achieving quality and practicality in universal NMT. While we prototype a high-quality universal translation system, our extensive empirical analysis exposes issues that need to be further addressed, and we suggest directions for future research.
@article{arxiv.1907.05019,
title = {Massively Multilingual Neural Machine Translation in the Wild: Findings and Challenges},
author = {Naveen Arivazhagan and Ankur Bapna and Orhan Firat and Dmitry Lepikhin and Melvin Johnson and Maxim Krikun and Mia Xu Chen and Yuan Cao and George Foster and Colin Cherry and Wolfgang Macherey and Zhifeng Chen and Yonghui Wu},
journal= {arXiv preprint arXiv:1907.05019},
year = {2019}
}