基于短语模式的文本情感分类研究  

Text Sentiment Classification Based on Phrase Patterns

作　　者：李钝[1] 曹付元[2] 曹元大[1] 万月亮[1] 

机构地区：[1]北京理工大学计算机科学技术学院 [2]山西大学计算机科学技术学院,太原030006

出　　处：《计算机科学》2008年第4期132-134,共3页Computer Science

基　　金：太原市科技局项目

摘　　要：文本倾向识别的研究在诸多领域有着广阔的发展前景,短语模式的文本情感分类是问答系统、信息安全、网上调查等研究的基础。本文从语言学角度出发,首先,分析词典中对词语义定义的特点,采用"情感倾向定义"权重优先的计算方法获得短语中各词的语义倾向度,然后分析短语中各词组合方式的特点,提出中心词概念来对各词的倾向性进行计算来识别短语的倾向性和倾向强度。实验表明,本文的方法对短语的倾向分类识别效果较好,可为更大粒度的文本倾向识别打好基础,具有一定的实用价值。The research of the text sentiment classification has the broad prospect and the related classification on phrases is the basic in QA system, information security and online investigation. Based on the linguistics, the paper analyzes the characteristics of the semantic definition in HowNet, calculates the orientation similarity of words in the phrase based on the sentiment weight priority and puts forward the concept of center word to calculate the orientation of the phrase according to the combination of the words in the phrase. The experiments show that it has better result in orientation recognition and some practice value that it can ground text orientation recognition in the research on larger granularities.

关 键 词：文本分类 情感倾向 语义倾向度 知网 短语结构 中心词 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP391[自动化与计算机技术—计算机科学与技术]