Microblog Sentiment Analysis And Monitoring System Based On Text Data Mining

In recent years, with the further development of the Internet and the maturation of mobile terminal technology, the social network has been integrated into people’s daily lives. More and more users express personal feelings, comments and other social hot spots on the social networks. Social network analysis has a vital role in public opinion analysis, having a widely spread concern from academia and industry. Microblog provides users a new network communication platform, to share and access anytime and anywhere. The number of users has increased steadily. Microblog containing different emotional characteristics tends to be sentiment analysis, which can dig out the microblog users’ point or idea on social events, also can be tracked on the development trend of events. Thus it is conducive to monitoring public opinion, rumor control and so on.In order to achieve emotional microblog analysis and public opinion analysis purposes, this article take fine-grained sentiment analysis, dividing the sentiment into happy, likeness, sadness, anger, disgust, fear, surprise these seven categories. By analyzing the microblog’s unique characteristics and short text emotion research status, this paper proposed a microblog emotional clustering method. Based on the integration of explicit and implicit characteristics, we develop a microblog sentiment analysis and monitoring system. The main contents include the following aspects:1) By analyzing the microblog own characteristics, we put forward the explicit emotional characteristics, including microblog smileys and emotional words. Then design and build the expression dictionary, included the network emotion dictionary words. We take explicit emotion frequent into mining properties, so the emotional characteristics feature items can get better preserve of the original microblog.2) Using both explicit and implicit emotion features, we propose a new systematic method for sentiment analysis. At first, the sentiment analysis dictionary, glossary of terms, as well as emoticons library from social networks, are initialized, and the text microblog frequent word sets are defined. Then according to the feature set of words, and maximum frequent itemsets, to obtain the initial clusters. Furthermore, for the microblog items overlap problem existing between multiple initial clusters, an efficient elimination method is proposed employing the extended membership degree of the short message semantic. Finally, the semantic similarity matrix is defined, based on which a hierarchical emotional clustering for microblogs is conducted. Taking the famous contest NLP&CC2013 in China as instance, the efficiency and priority of proposed method is proved by comparative experiments3) We build a complete microblog sentiment analysis and monitoring system, and through the sentiment analysis results on “Malaysia Airlines Flight 370” event’s microblog data sets, we verify the proposed method of microblog emotion analysis and monitoring capabilities, meanwhile, we visualize the clustering results.