基于句法路径的中文评论细粒度情感分析  

Fine-grained sentiment analysis of Chinese comments based on syntactic path

作　　者：胡征 陈尔希 曾献辉[1] 蔡一 HU Zheng;CHEN Er-xi;ZENG Xian-hui;CAI Yi(School of Information Science and Teehnology,Donghna University,Shanghai 201620,China)

机构地区：[1]东华大学信息科学与技术学院,上海201620

出　　处：《信息技术》2018年第9期116-120,共5页Information Technology

摘　　要：随着互联网的快速发展,篇章级的情感分析已不能满足应用需求,而对评论实体级或属性级的情感分析日益凸显价值。针对评价对象-情感词对提取过程存在的问题,文中采用实体、属性两级细粒度评价对象的表示方法,通过分析短语结构句法和构建句法路径库来识别情感单元,对于复杂句则结合中文句式分析来识别情感倾向。实验表明,文中方法对网络评论的细粒度对象表示和情感识别具有良好的效果。With the rapid development of Internet,chapter-level emotional analysis has failed to meet the application needs. And emotional analysis of the entity level or attribute level increasingly highlights its value. In order to solve problems during extraction processof object-emotion word pairs,this paper adopted two-level fine-grained representation of entity and attribute to express the affective units. The phrase structure syntax and syntactic path library were analyzed to deal with emotion units. For complex sentences,analysis of Chinese sentence pattern was used to identify affective tendencies. The experiments show that this method has a good effect on fine-grained object representation and emotion recognition of web comments.

关 键 词：细粒度对象 情感分析 短语结构句法 中文句式 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]