The Sentiment Of Classifition Research Of Micro-blog Based On Replying Messages

With the rapid development of the technologies of Web2.0, the micro-blog becomes the mostpopular social application after blog. The micro-blog is very convenience and free, a lot of peopleare attracted by it in short time and write a lot of messages in the platform of micro-blog everyday.These messages are all written by the people themselves, so the messages are very rich in theemotional content. They can be mined in order to know the attitudes to the hot issues and thegovernment’s policies, as a result we can offer the decision supports to the users, governments andcompanies. The paper uses the Sina micro-blog as the research object, and then the following jobsare done in order to propose a sentiment classification method to the Chinese micro-blog.Firstly, we collect some micro-blog messages from Sina, then the content length, the numberof sentences, the links, the labels, the emoticons and the pictures in the messages are analyzed fromthe statistical perspective. Analysis and comparison of these contents is important reference to thefeature extraction and algorithm design in this paper.Secondly, from the view of semantic rules we proposed a sentiment classification methodwhich is based on the micro-blog dictionary and replying messages of the micro-blog. According tothe construction of the domain emotional words dictionary, the method can do the sentimentclassification of the micro-blog combining with the replying message distribution of emotion. Wedesign a method which the replying messages can be measured, and the concept which is based onthe credibility of the replying messages is proposed in order to measure the authenticity of themessages.Thirdly, from the view of machine learning we proposed a sentiment classification methodwhich is based on the semantic features and replying messages of the micro-blog. The methodmakes use of the semantic features and the features of the Chinese micro-blog and the replyingmessages, uses the vector space model to represent the feature vector and also combines with theSVM classification method in order to realize the sentiment classification for the Chinesemicro-blog corpus.Finally, we make use of the Sina micro-blog’s corpus data to design some experiments in orderto verify whether the classification algorithm which we proposed above is correct. The results showthat the method which we present above is feasible and effective.