基于含权概念图的中文语句语义匹配方法的研究  

Research on Methods of Chinese Sentence Semantic Matching Based on Conceptual Graph with Weight

作　　者：刘培奇[1] 曹东川 刘阳[1] 

机构地区：[1]西安建筑科技大学信息与控制工程学院,陕西西安710055

出　　处：《微电子学与计算机》2015年第8期30-34,共5页Microelectronics & Computer

基　　金：国家自然科学基金项目(60673170);科技部科学技术支持项目(2008BAH37B05060);陕西省自然科学基金项目(2014JM2-6114);陕西省教育厅自然科学专项(12JK0743);教育部教改项目(2-3-ZXM-09)

摘　　要：提出了一种基于含权概念图的语义匹配方法.通过规则将中文语句信息转换为概念图,并且为各个概念节点赋予一个权值.通过语义库进行词语的相似度计算,计算出各个概念节点之间的相似度,再与节点权值进行加权计算,计算出整体概念图的相似度,达到中文信息语义匹配的目的.相对于传统的关键字语义匹配方法,所提出的方法对于同义异形词和同形异义词的匹配效果更好.实验结果表明,该方法能有效地实现中文语句的语义匹配,并且语义匹配准确率达到了60%.A semantic matching method is presented,which is based on conceptual graph with weight.Chinese statement information can transform conceptual graph according to some rules,and each concept nodes are assigned a weight.The word similarity is calaclated through semontatic library,then the similarity between the nodes of each concept is calculated with the node weights,calculated as the similarity of the overall conceptual graph,to achieve the purpose of Chinese Information semantic matching.Compared with traditional keyword semantic matching method,the proposed method is better for the match effect of Synonyms Word and Homographs Word.Experimental results show that the method can effectively achieve semantic matching of Chinese statement,and semantic matching accuracy reached 60%.

关 键 词：语义匹配 含权概念图 相似度计算 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]