Research On Chinese Microblog Sentiment Classification Based On Semantic Analysis

The emergence of Web2.0allows users to become creators and managers ofInternet information, and it thoroughly changed the one-way dissemination mode ofinformation on the Internet. Microblog as a typical Web2.0Internet application,since its introduction into the domestic, rapidly developed into a core social platformfor people to share and obtain information. Users express their views and emotionsthrough Microblog, to make the Internet produced a mass of text that containsemotional information. The research on Microblog sentiment analysis helps toachieve the work of Microblog regulation, finding and guiding public opinion, andbusiness competitive intelligence analysis. Compared with the traditional text,Microblog has the features of wide range topics, colloquial expression, and languagefragmentation, etc. The research on Microblog sentiment analysis will encountermore difficulties and challenges. Former domestic Chinese Microblog sentimentanalysis research is still at an early stage, there are a large number of researchquestions need to be addressed in depth. Therefore, the Microblog sentimentanalysis study has a high theoretical and practical value.Sina and Tencent Microblog as objects of the study, Chinese Microblogsentiment analysis-related techniques was introduced, including sentiment lexiconbuilding methods, Chinese Microblog sentiment analysis methods, and TextSentiment Analysis Experimental System (TSAES) design and implementation.In terms of sentiment lexicon building research, a semantic analysis-basedsentiment lexicon building method was proposed. Sentiment intensity strengthvalues of the word was automatically calculated by decomposing it into multipleEnglish semantic units. The proposed lexicon was applied to the task of sentimentanalysis, in which SVM (Support Vector Machine) was used to build the sentimentclassifier. Eexperimental results indicated that the built sentiment lexicon was moreeffective than the general polar sentiment lexicon. And also emoticons sentimentlexicon and the network language sentiment lexicon was built by drawing on theSO-PMI algorithm using corpus-based statistical methods.In terms of Microblog sentiment analysis techniques research, proposed asemantic analysis-Based Chinese Microblog sentiment classification method. Firstthe sentiment expression Binary Tree is constructed using the dependency parsingresult and the built sentiment lexicons. And then the sentiment category wasclassified according to the sentiment intensity strength value which was calculated by the established rules. Experimental results demonstrated the effectiveness of themethod, and the results also shown that the built emoticons sentiment lexicon andthe network language sentiment lexicon can effectively enhance the performance ofthe sentiment analysis system.In terms of Text Sentiment Analysis Experimental System design andimplementation research, A text (including Microblog) sentiment analysisexperimental system based on C/S and B/S hybrid architecture was built by usingthe proposed lexicons and sentiment analysis method to help to research on massivedata sentiment analysis. The system contains the tools of sentiment lexicon building,corpus management, and sentiment analysis visualization.