基于语义的词语相似度算法研究  

Study of Words Similarity Algorithm Based on Semantics

作　　者：刘燕美[1] 李华[1] 

机构地区：[1]重庆大学计算机学院,重庆400044

出　　处：《世界科技研究与发展》2012年第4期617-620,共4页World Sci-Tech R&D

基　　金：重庆市自然科学基金计划项目(2007BB2192);国家社科基金"十一五"规划教育学(ACA07004-08)资助

摘　　要：在《知网》的研究基础上提出了一种新的计算词语语义相似度的算法。在计算义原相似度时,深入研究了义原间的上下位关系和反义对义关系,考虑了义原间的共同节点数和语义距离,同时引用了《知网》最新发布的情感分析用词语集;此外,还分析了常见的弱义原和不带感情色彩的中性词,排除了它们对结果的干扰。实验结果表明,本算法使词语相似度计算结果更趋于合理,满意度更高。A new method of word semantic similarity computation based on HowNet is presented. Relationships of upper and lower and antonym relationships among sememes are deeply research. Common nodes and semantic distance between sememes are considered for computing sememe similarity, at the same time, emotional words published recently by HowNet are adopted, with weak sememes and neutral words ana- lyzed which have little emotion, excluding their effect on the accuracy of results. Experimental results show that results of word similarity computation are more reasonable and satisfied in this method.

关 键 词：知网 语义相似度 义原 对义关系 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]