一种本体概念的语义相似度计算方法  

A Semantic Similarity Measure between Ontological Concepts

作　　者：李文清[1] 孙新[1] 张常有[1,2] 冯烨[3] 

机构地区：[1]北京理工大学计算机学院,北京100081 [2]石家庄铁道大学信息科学技术学院,石家庄050043 [3]北京控制工程研究所,北京100190

出　　处：《自动化学报》2012年第2期229-235,共7页Acta Automatica Sinica

基　　金：国家自然科学基金(60873208);河北省自然科学基金(F2009000929)资助~~

摘　　要：概念语义相似度已广泛应用于Web服务发现、本体映射等领域,但现有的概念语义相似度计算方法对概念间语义相似程度的区分不够细致.本文从本体结构出发,首先提出了自底向上的本体概念出现概率计算方法,并在此基础上改进了基于节点信息量的概念语义相似性度量方法;然后又设计了基于边计算的本体概念语义相似度计算方法;最后对上述两种方法线性加权,提出了一种加权的本体概念语义相似度计算方法.实验结果表明该方法能进一步正确区分本体中父子概念及兄弟概念间的相似程度.Concept semantic similarity is wildly used in web service matchmaking, ontology mapping and so on. But the existing concepts semantic similarity measuring methods cannot distinguish the similarities further. So in this paper, we firstly propose a bottom-up concept probability computation method based on ontology structure, and based on this probability, we improve an information content based semantic similarity method. Then, we design an edge based concept semantic similarity method. Finally, we linearly combine the two previous semantic similarity methods to form a weighted one. Result shows that the weighted one can distinguish similarity between concept and its children, or between siblings.

关 键 词：本体 语义相似度 概念出现概率 信息量 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]