基于主题-对立情感依赖模型的虚假评论检测方法  

Detection Method of Fake Reviews Based on Topic-Opposite Sentiment Dependency Model

作　　者：汪建成[1] 严馨[1] 余正涛[1] 宋海霞[1] 石林宾[1] 

机构地区：[1]昆明理工大学智能信息处理重点实验室,云南昆明650500

出　　处：《山西大学学报（自然科学版）》2015年第1期31-38,共8页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(No.61175068;61462055);云南省自然科学基金重点项目(No.2013FA030);云南省软件工程重点实验室开放基金(No.2011SE14);教育部回国人员基金;云南省教育厅基金重大专项资助

摘　　要：根据评论内容的特性,提出了一种基于主题-对立情感依赖模型(topic-opposite sentiment dependency model,TOSDM)的虚假评论检测方法。首先构建TOSDM模型,利用该模型提取评论的主题信息以及主题对应的情感信息;然后结合评论的主题以及情感信息,分析并提取6维评论内容特征;最后利用这些评论内容特征,采用有监督学习的分类器对虚假评论进行检测。在大众点评网获取的2009-2012年的5个领域的评论数据集上进行了实验,实验表明,提取评论主题信息以及主题对立情感信息能够提高虚假评论的检测效果,TOSDM的虚假评论检测效果优于其他模型。Based on the content characteristics of the review,we proposed a method to detect fake reviews based on topic-opposite sentiment dependency model（TOSDM）.We construct TOSDM of reviews based on sentiment dependency,which is used to extract review topic information and their corresponding emotional information.Moreover,We analyze and extract 6types of review content features combined review topics and emotional information,and exploit all these review content features which were extracted in the back step to detect fake reviews using classifier based on supervised method.The experiments are completed with the reviews of 5fields from Dianping.com in 2009-2012.The experimental results show that extracting the reviews topic information and topic-opposite sentiment information can improve the detection effect for fake reviews,and TOSDM outperforms other generative models.

关 键 词：评论内容 虚假评论 评论挖掘 主题-对立情感依赖模型(TOSDM) 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]