基于神经网络的统计机器翻译的预调序模型

A Neural Pre-reordering Model for Statistical MT

长距离调序是统计机器翻译的一个主要挑战。之前的研究工作表明预调序是解决这个问题的一个可能的途径。在该工作中,我们沿着预调序这个研究方向,将神经网络建模结合到线性排序的框架之下,提出了一个基于神经网络的预调序模型。这个的预调序模型能够利用从海量未标注数据中抽取的句法和语意信息,从而更好的对不同语言之间的语序差异进行预测。我们在中文到英文以及日文到英文的机器翻译任务上进行了实验,实验结果表明了该方法的有效性。

Long distance reordering is a major challenge in statistical machine translation. Previous work has shown that pre-reordering is a promising way to tackle this problem. In this work, we extend this line of research and pro- pose a neural network based pre-reorder model, which integrates neural network modeling into a linear ordering framework. The neural network based model can leverage syntactic and semantic information extracted from unla- hel.ed data to predict the word order difference between languages. Experiments on Chinese-English, and Japanese- English machine translation tasks show the effectiveness of our approach.