Facebook FAIR's WMT19 News Translation Task Submission
This work addresses machine translation quality for news content, achieving state-of-the-art results but being incremental over prior methods.
The paper describes Facebook FAIR's submission to the WMT19 news translation task, which achieved first place in all four language directions (English <-> German and English <-> Russian) in human evaluation, with the English->German system outperforming other systems and human translations by improving 4.5 BLEU points over their previous submission.
This paper describes Facebook FAIR's submission to the WMT19 shared news translation task. We participate in two language pairs and four language directions, English <-> German and English <-> Russian. Following our submission from last year, our baseline systems are large BPE-based transformer models trained with the Fairseq sequence modeling toolkit which rely on sampled back-translations. This year we experiment with different bitext data filtering schemes, as well as with adding filtered back-translated data. We also ensemble and fine-tune our models on domain-specific data, then decode using noisy channel model reranking. Our submissions are ranked first in all four directions of the human evaluation campaign. On En->De, our system significantly outperforms other systems as well as human translations. This system improves upon our WMT'18 submission by 4.5 BLEU points.