一种基于概念关联式的词义消岐方法  

Word Sense Disambiguating Method Based on Concept Relativity

作　　者：缪建明[1] 张全[1] 

机构地区：[1]中国科学院声学研究所,北京100190

出　　处：《计算机科学》2010年第1期208-210,共3页Computer Science

基　　金：国家973项目"自然语言理解的交互引擎研究"(2004CB318104);国家科技支撑计划课题"搜索引擎中的语言翻译基础研究"(2007BAH05B02-05);中科院声学所知识创新工程项目"句群理解处理理论及其应用"(O654091431);中国科学院声学研究所"所长择优基金"(GS13SJJ04);中国科学院青年人才领域前沿项目(O754021432)资助

摘　　要：词义排岐是自然语言处理中最关键也是最困难的问题之一,至今仍没有得到完全有效的解决。在研究HNC表达汉语知识的基础上,提出了一种基于概念关联式的汉语词义消歧方法,用于处理汉语的歧义字段。该方法综合了词语概念的层次性、网络性、结构性特征,用一种统一的表示式来规范这类特征,解决了多个不同概念之间的知识关联表示问题。实验对20个汉语高频多义词进行了测试,平均正确率为94%,验证了该方法的有效性。As one of the most important and also the most difficult problem of Chinese information processing field, word sense disambiguation has not been entirely solved until now. On the basis of our research on HNC representing Chinese knowledge, this article proposed a word sense disambiguating method based on concept relativity. The method combines various features in concepts. The features include hierarchy, network and structure etc. The uniform representation formalizes the features. In this way, the problem of knowledge relationship expression among various concepts will be solved. 20 Chinese polysemous words were tested in our experiment. The result with average precision 87% shows that the method is effective.

关 键 词：词义消岐 概念关联式 HNC理论 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]