Deliberation network-based feature fusion for Chinese- English neural machine translation

Abstract

Neural machine translation (NMT) has achieved remarkable results in sentence-level translation, but the text problems of sentence-level translation, such as consistency and reference, are solved by using context information. Different from the previous methods using source context modeling, this paper proposes a novel Chinese-English neural machine translation that integrates target context information based on deliberation network. Specifically, with the help of the deliberating network, this paper makes a second translation of the source end of the text. The first translation is based on sentence level translation, and the second translation refers to the first translation of the whole text. The integration of domain knowledge into the translation model improves the effect of the translation model. The experimental results show that compared with the baseline model, the BLEU values of the Chinese-English and English-Chinese models are increased by 1.28 and 2.08.

Keywords

• chinese-english neural machine translation
• target context information
• deliberation network
• domain knowledge