基于概念间边权重的概念相似性计算方法  

Concept similarity computation method based on edge weighting between concepts

作　　者：冯永[1,2] 张洋[1,2] 

机构地区：[1]重庆大学计算机学院,重庆400030 [2]信息物理社会可信服务计算教育部重点实验室(重庆大学),重庆400030

出　　处：《计算机应用》2012年第1期202-205,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(61103114);重庆市高等教育教学改革研究重点项目(112023);"211工程"三期建设项目(S-10218);中央高校基本科研业务基金资助项目(CDJXS11181164)

摘　　要：介绍了传统的基于距离的相似度计算方法,针对其在距离计算中包含语义信息不充足的现状,提出了一种改进的使用WordNet的基于概念之间边的权重的相似性度量方法。该方法综合考虑了概念在词库中所处层次的深度和密度,即概念的语义丰富程度,设计了一种通用的概念语义相似性计算方法,该方法简化了传统语义相似性算法,并解决了语义相似性计算领域的相关问题。实验结果表明,所提方法在Rubenstein数据集上与人工判断有着0.910 9的相关性,与其他经典的相似性计算方法相比有着更高的准确性。The traditional distance-based similarity calculation method was described.Concerning that the method of distance calculation does not contain sufficient semantic information,this paper proposed an improved method which used WordNet and edge weighting information between the concepts to measure the similarity.It considered the level of depth and density of concepts in corpus,i.e.the semantic richness of concept.Using this method,the authors can solve the semantic similarity calculation issues and make the calculation of similarity among concepts easy.The experimental results show that,the proposed method has a 0.910 9 correlation with the benchmark data set-Rubenstein concept pairs.Compared with the classical method,the proposed method has higher accuracy.

关 键 词：概念相似度计算 WORDNET 边权重 语义信息 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]