为了提高基于短语的机器翻译系统的重排序能力,提出了一个基于源语言端的中心-修饰依存结构的重排序模型,并将该重排序模型以软约束的方式加入到机器翻译系统中.该排序模型提出了一种在机器翻译中应用句法树资源的方法,将句法树结构,通过将句法树映射成中心-修饰词的依存关系集合.该重排序模型在基于短语系统的默认参数设置下,显著地提升了系统的翻译质量.在系统原有的词汇化的重排序模型基础上,该重排序模型在翻译模型中融入了句法信息.实验结果显示,该模型可以明显地改善机器翻译系统的局部调序.
To enhance the reordering capacity of the phrase-based SMT (statistical machine translation), the studyleverages the head-modifier dependency structure on the source to model the reordering. The model is added tobaseline model in the form of soft-constraint way. The proposed model explores an approach to utilize theconstituent based parse tree that the parse tree is mapped into sets of head-modifier relationships. Experimentalresults show that this model improves the local reordering significantly.