Word Alignment Based on Multi-Grain Model

Word alignment plays a critical role in statistical machine translation (SMT) and cross-language information retrieval. Until now, most existing methods get the word alignment within the whole range of the sentence length. The alignment quality is unsatisfactory. In this paper, we propose a novel approach to word alignment based on multi-grain model (WAMG). We split a parallel sentence pair into blocks in different grain and get the word alignments within each corresponding block. Our approach is able to restrict the search space of word alignment in the relatively accurate local range and reduce the mapping error. The experiments have shown that our approach outperforms the traditional word alignment algorithm relatively by about 12% in AER and improves the performance of Chinese-to-English translation system relatively by about 2.8% in BLEU.