Exploring Hypotactic Structure for Chinese-English Machine Translation with a Structure-Aware Encoder-Decoder Neural Model
Guoyi Miao,Yufeng Chen,Mingtong Liu,Jinan Xu,Yujie Zhang,Wenhe Feng
DOI: https://doi.org/10.1587/transinf.2021edp7168
2022-01-01
IEICE Transactions on Information and Systems
Abstract:Translation of long and complex sentence has always been a challenge for machine translation. In recent years, neural machine translation (NMT) has achieved substantial progress in modeling the semantic connection between words in a sentence, but it is still insufficient in capturing discourse structure information between clauses within complex sentences, which often leads to poor discourse coherence when translating long and complex sentences. On the other hand, the hypotactic structure, a main component of the discourse structure, plays an important role in the coherence of discourse translation, but it is not specifically studied. To tackle this problem, we propose a novel Chinese-English NMT approach that incorporates the hypotactic structure knowledge of complex sentences. Specifically, we first annotate and build a hypotactic structure aligned parallel corpus to provide explicit hypotactic structure knowledge of complex sentences for NMT. Then we propose three hypotactic structure-aware NMT models with three different fusion strategies, including source-side fusion, target-side fusion, and both-side fusion, to integrate the annotated structure knowledge into NMT. Experimental results on WMT17, WMT18 and WMT19 Chinese-English translation tasks demonstrate that the proposed method can significantly improve the translation.