面向机器翻译系统句法分析器的研究与实现

Research on Parsing Technology in Machine Translation System

自然语言句法分析是机器翻译不可缺少的前期处理过程,文中总结了几代机器翻译系统中句法分析的经验,在最新研制的英汉双向机器翻译系统MTS2000中设计并实现了一个模块化的统计与规则相结合的句法分析模型。整个句法分析采用综合的策略,分别使用了隐马尔可夫方法、统计决策树方法、基于历史的句法分析等多种方法,并注意语义知识在句法分析中的应用。实验结果表明,模块化的句法分析器的设计方法,不论是对英语句法分析还是对汉语句法分析都是一种可取的方法。

Parsing is a central problem for many tasks in natural language processing,especially for machine translation.This paper presents the experiences in parsing for several machine translation systems,and proposes a modular parser using combination of statistic and rule methods.It integrates several strategies including Hide Markov method,statistic decision tree,history-based methods and employs semantic information in the algorithm.Experimental re-sults indicate that this approach is successful in both English and Chinese parsing.