网络论坛教学评论的自动情感分析方法——以湖南商学院枫华论坛为例  

On Sentiment Analysis of Campus Network Reviews about Education——a Case Study of Fenghua BBS of Hunan University of Commerce

作　　者：曾伏秋[1] 罗毅辉[1] 杨刚[1] 胡清洁[1] 

机构地区：[1]湖南商学院,湖南长沙410205

出　　处：《湖南商学院学报》2013年第1期106-110,共5页Journal of Hunan Business College

基　　金：湖南省科技厅计划项目(项目编号:No.2009CK3081)

摘　　要：校园网络论坛已经成为大学生实时发布各种观点和体会的公共平台。论坛产生的评论数据正以指数级的速度在增长,这些评论中包含了许多对教学管理有价值的信息,但是由于整个论坛的评论数量众多,很难用人工的方法抽取并且分析其中的教学评论情感信息,因此这些信息目前尚未得到管理者应有的关注。文章采用自动贝叶斯文本情感分类技术将校园网络论坛中的教学在线评论分成好评和差评情感信息,在湖南商学院枫华论坛上抽取的教学评论数据集上的分类准确度达到了87.3%,同时其结果的分布与人工标注结果的分布也不存在统计意义上的差异。该方法能够克服现有网上评教方法的一些缺陷,为教学管理部门实施动态教学管理提供了一种新的途径。Forum on a campus network provides a public and virtual platform for the university students to share their educational experience with other people. As a result, the exponential growth of reviews on forum occurs. Among these reviews, there is a huge amount of information about education available for teaching management. As today＇s students increasingly make their opinions available online, there have accumulated a huge amount of reviews on the forum. It is very difficult to deal with the huge amount of information by manually-operated methods; so much of it is ignored by people. This paper adopts Bayes text sentiment classification technology to address this problem by automatically classifying students＇ reviews about education into positive or negative opinions. Experimental results on the dataset obtained from Fenghua BBS of Hunan University of Commerce show that the accuracy of sentiment classification is about 87.5% and no statistically significant difference between the distributions in the classification result and those in artificial classification result. This work overcomes some weaknesses of traditional teaching evaluation and provides a new method by which the department can implement teaching management dynamically.

关 键 词：情感分析 教学评论 网上评教 贝叶斯分类器 

分 类 号：G642.0[文化科学—高等教育学]