Neural Machine Translation Leveraging Phrase-based Models in a Hybrid Search

摘要

In this paper, we introduce a hybrid search for attention-based neuralmachine translation (NMT). A target phrase learned with statistical MT modelsextends a hypothesis in the NMT beam search when the attention of the NMT modelfocuses on the source words translated by this phrase. Phrases added in thisway are scored with the NMT model, but also with SMT features includingphrase-level translation probabilities and a target language model.Experimental results on German->English news domain and English->Russiane-commerce domain translation tasks show that using phrase-based models in NMTsearch improves MT quality by up to 2.3% BLEU absolute as compared to a strongNMT baseline.