北京邮电大学学报

  • EI核心期刊

北京邮电大学学报 ›› 2013, Vol. 36 ›› Issue (2): 98-101.doi: 10.13190/jbupt.201302.98.wangt

• 研究报告 • 上一篇    下一篇

WordNet中的综合概念语义相似度计算方法

王桐1,吴吉义1,王磊2,徐贺3   

  1. 1. 哈尔滨工程大学 信息与通信工程学院, 哈尔滨 150001; 2. 杭州师范大学 电子商务与信息安全重点实验室, 杭州 310027; 3. 哈尔滨工程大学 机电工程学院, 哈尔滨 150001
  • 收稿日期:2012-10-03 修回日期:2012-11-06 出版日期:2013-04-30 发布日期:2013-03-25
  • 通讯作者: 王桐 E-mail:wangtong@hrbeu.edu.cn
  • 作者简介:王桐(1977-),男,副教授,E-mail:wangtong@hrbeu.edu.cn
  • 基金资助:

    国家自然科学基金项目(61102105,60775060);国家博士后科学基金项目(2008044840);教育部博士点基金项目(20102304120014,20102304110006);黑龙江省自然科学基金项目(F201029);浙江省自然科学基金项目(LQ12G02016,LZ12F02005)

Semantic Similarity Calculation Method of Comprehensive Concept in WordNet

WANG Tong1, WANG Lei1, WU Ji-yi2, XU He3   

  1. 1. College of Information and Communication Engineering, Harbin Engineering University, Harbin 150001, China;<br>2. Key Laboratory of E-Business and Information Security, Hangzhou Normal University, Hangzhou 310036, China;<br>3. College of Mechanics and Electrics Engineering, Harbin Engineering University, Harbin 150001, China
  • Received:2012-10-03 Revised:2012-11-06 Online:2013-04-30 Published:2013-03-25

摘要:

作为语义异构问题的基础,概念间语义相似度计算已成为研究热点,对此,提出一种基于WordNet的综合概念语义相似度计算方法. 该方法不仅集成了传统的基于语义距离的算法和基于信息内容的算法,而且引入了深度、密度因子和语义重合度来进行综合分析,并针对综合算法中权值难以确定的问题,引入主成分分析改进权值分配方法. 实验结果表明,改进后的方法计算的相似度与人工判断的相似度相关性较高,有效改善了概念语义相似度计算的准确性.

关键词: 概念语义相似度, WordNet, 主成分分析

Abstract:

As the basis of the semantic heterogeneity, the calculation of semantic similarity between concepts has become a hot topic. A calculation method based on the comprehensive concept of the semantic similarity of WordNet is presented. The method integrates traditional semantic distance-based algorithm, content-based algorithm, introduces the depth, density factor and semantic coincidence degree to conduct a comprehensive analysis. In order to determine the right weights in the synthesis algorithm, a principal component analysis is proposed to improve the weight allocation. Experiments show that the similarity of the proposed method has good correlation with similarity to the artificial one, thus the accuracy of the concept of semantic similarity calculation is improved effectively.

Key words: conceptual semantic similarity, WordNet, principal component analysis

中图分类号: