基于互信息的概念语义相似度算法  

Concept Semantic Similarity Algorithm based on Mutual Information

作　　者：种晓阳 周子力[1] 吴玲玲[1] 

机构地区：[1]曲阜师范大学物理工程学院,山东曲阜273165

出　　处：《通信技术》2016年第9期1175-1180,共6页Communications Technology

基　　金：山东省高校科技计划项目"无线多跳多蜂窝网络中协作信号处理和资源调度方法研究"(No.J15LN57);国家863计划项目"数据驱动的媒体内容动态自组织及封装技术研究"(No.2015AA015801)~~

摘　　要：随着信息技术的发展,概念语义相似度计算成为当前的一个研究热点。本文提出一种基于互信息的概念语义相似度算法,以IC概念信息内容模型为基础,将相邻概念的条件概率作为边的权值,将概念间的互信息作为语义相似度,不仅考虑概念所处分类树的深度、密度,而且将路径因素考虑在内。实验结果表明,与其他八种算法相比,该算法计算的相似度与人工判断的相似度相关度更高,相似度值更准确。With the development of information technology, the conceptual semantic similarity computation now becomes a hot research topic. A novel algorithm based on mutual information for calculating conceptual semantic similarity in WordNet is proposed. This algorithm, with IC（Information Content） model as the basis, the conditional probability of neighboring concept as the edge weighting information, and the mutual information as the semantic similarity, considers the depth and density of concept in the taxonomic tree of WordNet, even including the route factor. Experiments show that the similarity calculated by the proposed method has fairly good correlation with the similarity by artificial method, and the similarity is even more accurate.

关 键 词：概念语义相似度 条件概率 互信息 信息内容 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]