Lexical-based Reordering Model for Hierarchical Phrase-based Machine Translation

Reordering is a critical step for statistical machine translation. The hierarchical phrasebased model (HPB) based on a synchronous context-free grammar (SCFG) is prominent in solving global reorderings. However, the model is inadequate to supervise the reordering process so that sometimes phrases or words are reordered due to the wrong choice of translation rule. In order to solve this problem, we propose a novel lexical-based reordering model to evaluate the correctness of word order for each translation rule. Our approach employs the word alignment and translation information during the decoding process without causing too much extra computational consumption. Experimental results on the Chinese-to-English task showed that our method outperformed the baseline system in BLEU score significantly. Moreover, the translation results further proved the effectiveness of our approach.