Microblogging Sentiment Analysis For Automatic Extraction Of Key Technology Research Body

With the rapid development of newly Internet social applications, microblogs have been widely applied in281million users and its utilization ratio in2014is45.5%. Every day, millions of users share their opinion and sentiment on different topics. Automatically analyzing the user’s sentimental trends and scientifically determining the tendency of the whole microblog community under a certain topic, have become the basic scientific problem on the domain of microblog computation and public sentimental analysis.However, traditional sentiment analysis mainly deals with classicallong text. Since microblog contents are normally short and irregular and with the characteristics of scattered and personalized, traditional sentiment analysis algorithms cannotwork well and meet the practical requirementsduo to its essential defects and inefficiency. It is a simple and effective method to use emotion dictionary to parse user-generated content. However, due to the relative small size of the emotional dictionary, and an endless stream of new terms in Network, varieties of informal expressions, it is time-consuming to add new terms manually to the dictionary, but existing sentiment analysis method is more complex, and closely involved with field knowledge. To solve the above problems, this paper proposes an automatic emotion weight (AEW) algorithm to mine the potential emotional words and calculate the emotion weight, which is independent to the applications area and has good scalability. The method is based on Bayesian theory and big data mining technique, able to recognize unknown emotion words, judge the sentiment polarity and emotional orientation degree according to the value of its emotion weight. It effectively extends the emotion dictionary, rich its fine use, and automatically realizes the mining of emotion lexicon. Besides that, this paper also realizes the recognition of emotional subject attribute, including opinion sentence identification, sentiment object extraction and sentimental tendency adjustment, thus completing the ontology extraction of sentiment analysis.The paper makes the verification based on the above research, and at the same time, in order to verify this method can realize cross domain, this paper conduct three empirical analysis of integrated shopping mall JD.com, douban.com and dianping.com, the precision reached above90%. Preliminary experimental results show that the proposed model is a validated and practical strategy. It demonstrates that the AEW is fundamental algorithm for sentiment analysis of various Internet applications not only Weibo.