Research On Hot Event Sentiment Analysis For Topic Microblog

With the rapid development of microblog in China,domestic consensus almost often become a hot topic in the microblog, and more and more users participate in the discussion, microblog has gradually become an important real-time information sources and media center. After the hot events occurred, people often obtain the event information through the microblog in the first time, and discuss in the microblog, which produces some hot topics. Through mining the information of these topics,and analysising public opinion,we can find the degree of people’s concern and support about the event.It is important for the government to control and monitor the emergencies of people’s psychology.In order to obtain the evolution of public opinion about hot topic event in the microblog, this thesis employed emotion and tracked public opinion to analyze the corpus of popular events topic "The female driver was beaten by the male driver" in the Sina microblog.This thesis studied the features of hot topic microblog based on semantic lexicon and CRF model, compared the effect of polarity classification with different combinations of features through multiple sets of experiments. Then,this thesis mined opinion targets of microblog based on the rules and syntactic dependency analysis. In face of diversity of opinion targets extraction, the thesis clustered the extraction results to find more user’s focus points on hot events, to finally realize the evolution analysis of the public opinion on the hot topic microblog. The research mainly includes the following aspects:(1)The features of Microblog text are studied,from these phenomena we can analyze various features which is useful for classification, and describes the existing state of affairs of sentiment analysis research, and domestic and international evaluation conference on sentiment analysis.(2)We build a text-Chinese microblog semantic lexicon,which is based on the HowNet semantic lexicon, NTUSD and Dalian University emotional vocabulary ontology,and we have expanded it.(3)Based on the sentiment lexicon, we used CRF model with various feature combinations to analysis emotion of microblog,we choose four kinds of text features, including word,speech, emotional words and negative words,and design different combinations of features to make the effect optimal through multiple sets of experiments.(4)Different users often have different focus points and different opinion targets to the same event, so this thesis introduced the opinion target extraction strategy. The potential opinion target of the thesis is defined as a noun or noun phrase, and topic labels, based on the definition of Chinese noun phrase extraction potential evaluation object. After using rules and syntax based on dependency analysis method to extract <emotional word, opinion target>, and extract targets. Afterwards, this thesis clustered the extraction results.Finally, this thesis analysis the trend of public emotion variation.(5)In this thesis, the experimental data set is from microblog review about hot topic event "The female driver was beaten by the male driver" on the Sina microblog, which are climbed by the Sina API.