基于统计语言模型的译文选择方法

Language translation selection method based on word model

机器翻译是把一种自然语言转变成另一种自然语言的过程。其中 ,源语言词汇的译文选择直接决定机器翻译的质量 ,是人们关注的难题。本方法以信息论和计算语言学为理论基础 ,充分利用了日语分析生成的句子结构信息和语言模型提供的统计信息 ,来解决机器翻译中歧义词的译文选择问题。将该方法用于日汉机器翻译系统的译文选择实验 ,具有良好的译文选择性能。它的特点在于利用了日语分析的句法信息 ,使译文选词的质量得到明显改善。

Machine translation is the process of transforming one kind of natural language to another kind of natural language. Translation selection is a key problem and remains an open field. Based on Information theory and computational linguistics, our method tries to resolve the word selection by the information of Japanese word analysis and language model. The translation selection experience for a Japanese-Chinese machine translation system proves the satisfactory performance of this method. Its characteristic lies in that the method improves distinctly the quality of word selection in Japanese - Chinese translation taking advantage of Japanese analysis syntax information.