基于HowNet的句子褒贬倾向性研究  

Sentence orientation research based on HowNet

作　　者：熊德兰[1] 程菊明[1] 田胜利[1] 

机构地区：[1]许昌学院计算机科学与技术学院,河南许昌461000

出　　处：《计算机工程与应用》2008年第22期143-145,共3页Computer Engineering and Applications

基　　金：河南省教育厅自然科学基金项目(No2008B520031);许昌市科技局科技发展计划项目(No07020062)

摘　　要：文本倾向性识别在信息过滤、自动文摘、文本分类等领域有广泛的应用前景。句子倾向性研究是文本倾向性识别的基础,结合句法分析结果和词语语义倾向性可以衡量句子褒贬倾向性。以HowNet的词汇语义相似度计算为基础,提出了基于的语义距离和语法距离的句子褒贬倾向性计算方法。大量语句实验表明,该方法的计算结果与人工判别结果更接近。Text orientation recognition has broad application in some field such as information filtering,automatic abstracting and text classification.The sentence tendentiousness research is the text tendentiousness recognition foundation,the authors measure appraisal tendentious of a sentence considering of the syntax analysis result and words semantics orientation.Based on the HowNet semantic similarity calculation,the paper points a new method to compute appraisial orientation of sentence in the base of syntax and semantic distance.Experiments in some statements show that results of the method are closer to manual evaluation.

关 键 词：语义相似度 语法分析树 语义距离 语法距离 倾向性识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]