基于HowNet的信息量计算语义相似度算法  

Method of Information Content Evaluating Semantic Similarity on HowNet

作　　者：游彬[1] 严岳松[1] 孙英阁[2] 刘靖 

机构地区：[1]海军指挥学院信息战研究系 [2]中国人民解放军海军92665部队 [3]海军指挥自动化站

出　　处：《计算机系统应用》2013年第1期129-133,共5页Computer Systems & Applications

摘　　要：语义相似度计算的应用范围广泛,从心理学、语言学、认知科学到人工智能都有其应用.提出了仅依赖于知网(HowNet)的信息量计算来估计两个词汇间的语义相似度.经实验证明,相比于传统的基于词网(WordNet)和大型语料库的计算信息量来估计语义相似度的算法,本文的算法更容易计算,并更接近于人工的语义相似度判断.Evaluating Semantic similarity is widely used in areas range from Psychology, Linguistics, Cognitive Science to Artificial Intelligence. This paper means to the merely use of HowNet to evaluate Information Content as the semantic similarity of two terms or word senses. While the conventional ways of measuring the IC of word senses must depend on both an ontology like WordNet and a large corpus, the experiment proves that the semantic similarity measured in this method is easier to calculate and more closely with human judgments, as HowNet has an elaborate way to represent descriptive object.

关 键 词：语义相似度 HOWNET 信息量 语义距离 相似度计算 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]