Modeling Coherence for Discourse Neural Machine Translation

  • Hao Xiong Baidu, Inc.
  • Zhongjun He Baidu, Inc.
  • Hua Wu Baidu, Inc.
  • Haifeng Wang Baidu, Inc.

Abstract

Discourse coherence plays an important role in the translation of one text. However, the previous reported models most focus on improving performance over individual sentence while ignoring cross-sentence links and dependencies, which affects the coherence of the text. In this paper, we propose to use discourse context and reward to refine the translation quality from the discourse perspective. In particular, we generate the translation of individual sentences at first. Next, we deliberate the preliminary produced translations, and train the model to learn the policy that produces discourse coherent text by a reward teacher. Practical results on multiple discourse test datasets indicate that our model significantly improves the translation quality over the state-of-the-art baseline system by +1.23 BLEU score. Moreover, our model generates more discourse coherent text and obtains +2.2 BLEU improvements when evaluated by discourse metrics.

Published
2019-07-17
Section
AAAI Technical Track: Natural Language Processing