Facebook FAIR’s WMT19 News Translation Task Submission

TLDR

Following our submission from last year, our baseline systems are large BPE‑based transformer models trained with the FAIRSEQ sequence modeling toolkit. This paper describes Facebook FAIR’s submission to the WMT19 shared news translation task and experiments with different bitext data filtering schemes and filtered back‑translated data. We participate in four language directions (English↔German and English↔Russian), ensemble and fine‑tune models on domain‑specific data, and decode using noisy channel model reranking. Our system improves on our previous system’s performance by 4.5 BLEU points and achieves the best case‑sensitive BLEU score for English→Russian.

Abstract

This paper describes Facebook FAIR’s submission to the WMT19 shared news translation task. We participate in four language directions, English <-> German and English <-> Russian in both directions. Following our submission from last year, our baseline systems are large BPE-based transformer models trained with the FAIRSEQ sequence modeling toolkit. 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 system improves on our previous system’s performance by 4.5 BLEU points and achieves the best case-sensitive BLEU score for the translation direction English→Russian.