Research On Emotional Dictionary Based Sentiment Analysis Of Chinese MicroBlog

The arrival of Web2.0era, changing the mode of transmission of the interactionbetween traditional media and human information，microblog is a good example. Moreand more people are actively expressing their opinions and sharing information andfeelings on microblog. Due to its brief writing, convenient publishing and real-timeinteracting, microblog becomes very popular. Now many users through microblog toexpress their feelings or express their views on specific things. Analyzing sentimenthidden in these information can benefit microblog marketing, monitoring publicopinions and brand survey. Currently, Chinese microblog sentiment analysis is still atthe initial stage.Sentiment analysis is mainly to identify the sentiment tendencies of microblog,classifing users’s sentiment into positive and negative. Firstly, we crawl microblog fromSina Weibo and conduct some preprocessing on the corpus. Then, use Semi-supervisedapproaches to sample microblog containing specific expression and build the initialtraining set. Secondly, combining manual sentiment dictionary and commonly usednetwork words on the web, we use the extended Semantic Orientation Pointwise MutualInformation (SO-PMI) and Distributed representations of words to performexperiments on training set and build an sentiment dictionary for microblog. Finally,we use the sentiment words in the microblog as feature, train a sentiment tendenciesmodel used SVM method and obtain the microblog polarity, A system of sentimenttendencies analysis for microblog is implemented in this paper.