计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (22): 143-145.DOI: 10.3778/j.issn.1002-8331.2008.22.042

• 数据库、信号与信息处理 • 上一篇    下一篇

基于HowNet的句子褒贬倾向性研究

熊德兰,程菊明,田胜利   

  1. 许昌学院 计算机科学与技术学院,河南 许昌 461000
  • 收稿日期:2007-10-11 修回日期:2008-01-10 出版日期:2008-07-11 发布日期:2008-07-11
  • 通讯作者: 熊德兰

Sentence orientation research based on HowNet

XIONG De-lan,CHENG Ju-ming,TIAN Sheng-li   

  1. Department of Computer Science and Technology,Xuchang University,Xuchang,Henan 461000,China
  • Received:2007-10-11 Revised:2008-01-10 Online:2008-07-11 Published:2008-07-11
  • Contact: XIONG De-lan

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。

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

Abstract: 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.

Key words: semantic similarity, syntactic tree, semantic distance, grammar distance, tendentious recognition