Research Of Opinion Mining For Campus Public Opinion Analysis

As an important part of the Internet, the construction of the campus BBS completes the new mode of digital campus, and become a platform to obtain information and communicate between teachers and students. But as the spread and gradually thorough of opinions, the initial personal opinions may be converted to most people’s group consciousness, and finally formed a certain scale of the campus public opinion. Therefore, it’s necessary to deeply analyze the text sentiment orientation, accurately identify the opinion leaders of campus, and build a powerful public opinion system for campus BBS.For text sentiment orientation analysis, this paper combines with the common emotional expression used in campus BBS, and proposes a text opinion mining model based on emotional expression. The main study involves the following three modules, text preprocessing, text feature extraction and text sentiment orientation analysis based on SVM. In the text preprocessing module, this paper studies the existing emotional vocabulary resources, summarizes and organizes the auxiliary dictionaries including the user dictionary, stop words dictionary, negative word dictionary and emotional word dictionary. Taking advantage of auxiliary dictionaries, it’s convenient to improve the segmentation accuracy, meanwhile, reduce some noise vocabulary with high frequency, which further refine the result of the text preprocessing. In the text feature extraction module, this paper considers the impact of Bi-gram and sentiment phrases, punctuation symbols, emotion icons for emotional expression.For identification of opinion leader, this paper propose a hybrid data mining approach based on user feature and interaction network to identify opinion leader in campus BBS. This study includes three parts, a way to analyze users’ authority, activity and influence, a way to consider the orientation of sentiment in interaction network and a combined method based on HITS algorithm for identifying campus opinion leaders.Finally, based on the experimental data crawled from campus BBS, this paper designs and analyzes experiments for text sentiment analysis and identification of opinion leader. Comparative experimental results show that the proposed methods can provide an effective mining of the user sentiment and a better rate of recognition.