改进的基于知网词汇语义褒贬倾向性计算  

Improved lexical semantic tendentiousness recognition computing

作　　者：杨昱昺[1] 吴贤伟[1] 

机构地区：[1]宁波大红鹰学院电子信息学院,浙江宁波315175

出　　处：《计算机工程与应用》2009年第21期91-93,108,共4页Computer Engineering and Applications

基　　金：浙江省教育厅科研项目(No.20071322)

摘　　要：词汇语义褒贬倾向性研究是句子褒贬倾向性识别的基础,而句子褒贬倾向性识别又是文本倾向性识别和篇章结构褒贬倾向性识别的基础。以《知网》的词汇语义相似度计算为基础,针对目前采用计算基准词对与词汇相似度的方法识别词汇褒贬倾向性理论,从褒贬基准词和计算公式入手,提出了改进办法。实验证明,在同样基准词对下,准确率得到了很大的提高,达到98.94%,具有实际应用价值。Lexical semantic tendentiousness recognition computing is the base of the sentence tendentiousness,and the sentence tendentiousness recognition is the text tendentiousness recognition and the chapter structure tendentiousness recognition foundation. Based on HowNet lexical semantic similarity computing,according to the current vocabulary appraise tendentious theory used by the method of computing similarity between words and benchmark words,from appraise benchmark words and computing formula, the paper puts forward an improved method.With experiment validation,in the same pair of benchmark words,accuracy rate has greatly been improved,arriving at 98.94%,there is some value in practical application.

关 键 词：语义相似度 倾向性识别 知网 褒贬基准词 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]