Research On Sentiment Classification Oriented To Internet Information

Along with the development of Internet, the influence of the Internet Media is more and more strong. Huge number of people exchanging information, sharing their opinion on the Internet and letting off their emotion produce the public opinion. Because of more and more Internet public events, some national sectors have decide to enhance the power to gather and analyze the Internet Information. Public opinion system gathers and analyses the Internet Information by computer and can help the leaders make the decision effectively.This paper studies one of the key technologies in building the public sentiment system—the technology of emotion tendency of text. On the basis of current methods, we improve the effective of classifier by combing the classifiers. And then build the public sentiment system on the research of the related technologies. The work included is as followed:Firstly, this paper summaries the commonly used classifiers, and test them on multi-domain corpus. The public sentiment system may gather information of every domain, so test the effect of those classifiers on multi-domain corpus is needed.Secondly, this paper a new method of classifying the text emotion tendency based on reliability. This method use SVM、ME and KNN as based classifiers. Then choose the proper classifiers to classify the texts based on reliability of KNN.Thirdly, on the basis of studying the related technologies of public sentiment system, we build a public sentiment. By the effect of using of the system in working, we prove the design plan can support some national sectors well.