Enterprise Information Classification And Sentiment Analysis For Sina Micro-blog

Sina micro-blog has become the largest information sources in the present society.People can find what they want from the micro-blog, and know about what they want aboutany thing. It brings the enterprise with potential business opportunities. The enterprise cansearch related news in micro-blog information, and look for reflects to a new product of theenterprise. Then the enterprise understands what the types of products people like, and graspwhat the enterprise in people eyes. However, the information searched from the micro-blogdirectly is ambiguous and useless, because the information contains a lot of informationirrelevant with enterprises. Therefore, we must determine the information which is related tothe enterprise and which is not related to. Of course, what the enterprises pay most attentionto is not how many truly relevant with enterprise in a micro-blog. What they care about isusers’ emotional expression of the enterprise products, image, cultural and other aspects, andthe proportion in these micro-blogs.In view of the problems of the enterprise information analysis and micro-bloginformation characteristics, it is very necessary for the studies on the micro-blog informationclassification and sentiment analysis. Similarly, for enterprises, our research has veryimportant practical significance.The specific work is as follows:(1) The specific steps and methods of the text classification and text sentiment analysisare summarized and analyzed deeply, and the advantages and disadvantages of variousmethods are discussed in details;(2) We combine the classification and sentiment analysis technologies creatively,providing a feasible implementation method for enterprise information mining. And thismethod has been applied to the Sina micro-blog;(3) In the process of enterprise information classification, we propose a method ofclassifier based on real-time information flow learning, in order to solve the limitation oftraditional classifier. The experimental results show the method is efficient and accuracy.(4) In the process of the enterprise information sentiment analysis, in view of thecharacteristics of information in Sina micro-blog, we construct an affective dictionary withmicro-blog feature, providing a solid foundation for sentiment analysis. Moreover, in orderto solve the problem of the strength of sentiment influence for modifying adverbs and the negative adverbs to the emotional center word, we introduce a concept of distance betweenthe negative adverbs to the emotional center of word, enhancing the strength calculation ofsentiment accuracy. Experimental results show the necessity of the concept of distance.