会议论文

Text similarity detection algorithm based on component relation map

中文

2015

482-493

第十二届全国信息隐藏暨多媒体信息安全学术大会

第十二届全国信息隐藏暨多媒体信息安全学术大会论文集

2015-03-28

武汉

中国电子学会通信学分会<&wdkj&>北京电子技术应用研究所

本校为第一机构

提出一种新的基于部件关联图的文本相似度检测算法.根据汉字数学表达式理论将文本汉字拆分成部件,统计每个部件出现的频次并构建部件直方图;接着,构造出每个部件在文本中与之相关联的部件相互关联组成汉字的关系构建部件关联图;最后对部件关联图进行图匹配,得到关联图的相似度,最终得出文本的相似度值.实验结果表明CRM-TSD算法具有较好的准确率、召回率和F值;在与余弦定理和Jaccard系数方法的对比实验中表明,CRM-TSD算法在综合评价指标F-度量值上也优于这两种方法.

　　The traditional text similarity detection method almost used VSM to represent the text, but it is high dimensional and sparse.So we proposed a new text similarity detection algorithm using component relation map (CRM-TSD).Firstly, split the Chinese characters into components by the structure of the Chinese characters and the Mathematical Expression of Chinese Characters and constructed the component histogram map after calculated the frequency of each component.Secondly, built the component relation maps according to the relation between co...