常识推理在机器翻译词义排歧中的应用

The Application of General Knowledge Reasoning for Word Meaning Disambiguation in Computer Translation

本文针对词义排歧提出了一种利用语言常识进行逻辑推理以筛选出合理词义的形式化标准和方法,将人们在翻译过程中排歧时所进行的逻辑推理归结为一种机械的集合运算,使之易于机器操作.并在此基础上,提出一种利用义项多元组引入词义的语境相关限制信息、以改进现有电子词典使其更加有利于排歧的方案,从方向上指明了这种词典知识获取的途径.

The formalized criteria and method of selecting proper word meaning based on logical reasoning with general language knowledge are presented in this paper. By the method, the logical reasoning in human's activities of word meaning disambiguation is transformed into a mechanical set operation which is easy to perform for computer. On this basis, the paper provides a scheme to improve existing electronic dictionaries for more effective word disambiguation——a dictionary based on 'multi-place sequence of word meaning item', thus clarifies the orientation in getting the knowledge of those dictionaries.