摘要微博、论坛等自媒体平台的兴起与发展,改变了传统的信息由媒体流向个人的单向传播方式,如今每个人都既是信息的接收者,也是信息的发布者,而这也使得人们每天所面对的网络信息量呈爆炸式增长。因此如何从海量信息中快速把握关键信息,尤其对于政府和企业等机构来说,如何从大量网民评论中迅速把握主流观点,进行及时准确的舆论引导,就成为目前研究的热点问题。本文试图利用LDA主题模型对与某一舆情事件相关的微博及评论进行主题发现及文本聚类,并从社会网络结构的视角实现对“用户-所属观点主题”2-模网络的可视化及对观点主题演化过程的展示,主要的研究内容如下:84058
(1)通过舆情信息内容、用户关系、用户行为三个方面的四个维度(时间维/用户维/内容维/观点维)的关联,构建网络舆情观点主题发现方法体系。
(2)选取新浪微博平台上关于“双汇进口美国猪肉”事件的微博及评论为研究对象,构建以微博用户、微博及评论为节点,回复和点赞关系为连线的“用户-微博”复杂社会网络及其子网络,并利用社会网络分析软件Pajek实现了对复杂网络及其子网络的可视化。
(3)对与“双汇进口美国猪肉”这一事件相关的微博和评论数据进行数据预处理,使用JAVA版LDA模型,对每日微博数据集进行主题抽取。对抽取出的主题特征词项进行人工归纳、总结,得到每日微博观点主题。
(4)采用LDA模型直接聚类的方法,对微博及评论数据集进行文本聚类,得到每条微博最大可能所属观点。
(5)在聚类结果的基础上构建“用户-所属观点”2-模网络,并利用社会网络分析软件Pajek实现该网络的可视化。从对可视化效果图的观察及社会网络入度指标的测度结果中可以得到舆情事件参与主体所持有的主要观点,及观点主题随时间推移的演化情况。
实验证明利用LDA模型和人工总结可以得到舆情事件参与主体的主要观点,且结果较为准确,但仅利用LDA模型进行文本聚类的效果并不理想。同时,基于社会网络视角对舆情观点主题进行演化分析是可行的。
关键词:网络舆情,社会网络,LDA模型,主题发现,文本聚类
毕 业 论 文 外 文 摘 要
Title Research on the Topic Discovery of Online Public Opinion Based on Social Network Analysis
Abstract The development of micro-blog and forum has changed the traditional way of information spreading, which is spread from the media to the public。 Today, everyone is both recipient of information, but also the publisher of information。 While it also makes the amount of network information that people face exploding。 Therefore, how to grasp the key information from the mass of information quickly and how to grasp the mainstream view from a large number of Internet users’ review quickly, so that the government, enterprises and other organizations can lead the media guide timely and accurately, has become a hot topic of current research。 This paper attempts to use LDA topic model for text clustering of network reviews, and to achieve visual impressions or micro-blog view from the perspective of social network structure。 The main contents of this paper are as follows:
(1)Construct method system of topic discovery of the network public opinion based on three aspects of information content of public opinion, the relationship between the users and the users’ behavior in four dimensions(time dimension/user dimension/content dimension/view dimension)。
(2)Select the micro-blog and comments about incident named “Shuanghui imports the pork from US” from Sina micro-blog platform as an object of this study, in order to build a social public opinion network which takes the micro-blog users, micro-blog and comments as the nodes and takes the relationship of commenting and thumbing as connection。 This social public opinion network shows the relation between micro-blog, micro-blog users’ comments and thumbs, from the perspective of the social network structure。