文本倾向性分析综述  

Survey on sentiment orientation analysis of texts

作　　者：厉小军[1] 戴霖[1] 施寒潇[1] 黄琦[2] 

机构地区：[1]浙江工商大学计算机与信息工程学院,浙江杭州310018 [2]浙江大学计算机科学与技术学院,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2011年第7期1167-1174,1186,共9页Journal of Zhejiang University：Engineering Science

基　　金：浙江省重大科技专项资助项目(2008C13082);浙江省自然科学基金资助项目(Y1090688);中央高校基本科研业务费专项资金资助项目(2009QNA5025)

摘　　要：介绍文本倾向性分析的基本流程,从主观性文本识别、文本倾向性分析方法、现有系统及评测方法、语料库建设4个方面对现有文本倾向性分析技术进行介绍和概括.综述了文本倾向性分析的3类研究方法:简单统计方法、机器学习方法和细粒度情感相关性分析方法,分析这3类研究方法的特点,从算法复杂性、效率和适用范围等方面比较各自的优缺点.概括现有研究的成就和不足,从基础性问题、具体应用的实现方法2个方面提出研究的前景.The basic flow of sentiment orientation analysis of texts was introduced,and the primary four aspects of current interesting researches were presented： subjectivity text recognition,sentiment orientation analysis method of texts,existing systems and evaluation methods,construction of corpus.Then three methods and their characteristics were summarized,i.e.simple statistics,machine learning and fine-grained sentiment relative analysis method.Merits and demerits of methods were analyzed from complexity,efficiency and applicable scope.Finally,the current achievements and shortages were summarized,and forecasted research perspectives were proposed including basic problem and implementation method of specific application.

关 键 词：文本倾向性分析 情感语料库 主观性文本识别 意见挖掘 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP391[自动化与计算机技术—控制科学与工程]