基于HNC理论的词语相似度计算  

A New Measure of Semantic Similarity Based on Hierarchical Network of Concepts

作　　者：吴佐衍 王宇[1] 

机构地区：[1]大连理工大学管理科学与工程学院,辽宁大连116024

出　　处：《中文信息学报》2014年第2期37-43,50,共8页Journal of Chinese Information Processing

基　　金：教育部人文社科研究规划项目(09YJA870005);国家自然科学基金重大项目子课题(70890083)

摘　　要：该文运用自然语言处理的概念层次网络(Hierarchical Network of Concepts,HNC)理论提出了一种词语相似度计算方法。该方法利用HNC理论词汇层面联想的概念表述体系,根据HNC映射符号的编码规则和符号映射理论,综合概念内涵、概念外部特征、概念类别和组合符号来计算词语的相似度,并与基于知网的词语相似度算法和人工的主观判断的相似度进行了比较分析。实验结果表明,该方法能够较好地反映词语之间的语义差别,与人的直观判断基本一致,是一种有效可行的方法。A new measure based on Hierarchical Network of Concepts（HNC） theory is put forward to compute the semantic similarityin natural language. Based on the coding rules and the map theory included in the concept expres- sion form in the vocabulary relation level of HNC, the method integrates the concept of connotation, outward fea- tures, classification and combination of symbol to calculate semantic similarity. This method is compared with the current popular similarity methods based onHowNetaccording to the subjective judgment of human. Experiment showsthat the method has a good performance, which can distinguish the differences between different words more accurately.

关 键 词：概念层次网络 语义相似度 中文信息处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]