Phrase reordering models integrating words, phrases and language models for statistical machine translation

在这个项目中，我们开发了一种精确的长距离短语重排规则提取方法，这些规则是由介词等功能词词汇化的。由于虚词的模式决定了句子的结构，因此规则可以在结构的基础上正确地对短语进行排序。在日英翻译实验中，我们使用提出的规则集提高了翻译系统的性能。

In this project, we developed an extraction method for accurate and long-distance phrase reordering rules that are lexicalized by function words such as prepositions. Since patterns of function words determine the structure of a sentence, the rules can correctly reorder phrases on the basis of the structure. In Japanese-English translation experiments, we improved the performance of our translation system using the proposed rule set.