Social Media-Oriented College Internet Public Opinion Analysis System

University campus network public opinion to the ideological and political education of colleges and universities put forward new challenges, mainly related to campus security, school reputation, the interests of teachers and students, management decision-making and other content. In the development of the mobile Internet and promotion, the network public opinion mainly formed student website, microblogging, micro channel, almost carrier, the media has a huge number of users, user participation degree is high, the rapid dissemination of information, more and more students like to use the instant messaging software to share information, How to use the network and other scientific and technical means to monitor and guide the campus network public opinion, to further promote the construction of a harmonious campus, is the current problem to be solved.In this paper, based on weibo(http://weibo.com/) and zhihu(http://zhihu.com/) data, rely on the natural language processing technology and machine translation model design and implement the social media network public opinion analysis system.and this paper mainly has the following work:We use the Labeled-LDA topic model mining the large-scale micro-blog text, and get the interest model of each user. For the text classification task, we proposed an efficient Boosting Decision Tree algorithm, by the experiments we improved that the GBDT algorithm proposed in this paper has a great improvement on the speed and precision of SVM.For the public opinion analysis tasks. We use the machine translation of word alignment training based on segmentation algorithm to build the document title translation,and then used IBM-Model1 training translation models to constructed the semantic relationships between documents and keywords and then complement the public opinion keyword extraction task. The effectiveness of the method is demonstrated by comparing the experimental results with the existing field results and algorithms.Finally, based on the Labeled-LDA topic model and public opinion keyword extraction algorithm, we design and implementate a practical social media-oriented college analysis system.