The Research On Emotional Orientation Of Web Comments Based On Latent Semantics

With the popularity of the Internet,the way in which people get information is gradually changing from newspapers,radios,televisions and other traditional media to the Internet.More and more people are keen to exchange information via forum,blog and microblogging etc.Nowadays,the Internet has become the main carrier of the information dissemination.However,the openness,pluralism and real-time char-acteristics of the Internet,at a certain level,make the network become a breeding ground for the proliferation of rumors.The web information content skyrockets along with the surge in the number of Internet users.Therefore,how to monitor and perdict public opinion effectively is very important.The technology of emotional orientation analysis can dig out emotional tendency of the topic from the mass of Web comments,and find the hotspot of public opinion and predict its tendency quickly.With this technology,the monitoring and predicting of public opinion can be efficient.However,it also has some problems for emotional orientation analysis,such as low rate of classification accuracy and restricted of techn-ology's application.Thence,this paper will focuses on the problem of sentiment cla-ssification and application of key technologies.The main contents are as follows:1)To deal with the problem that existing sentiment classification methods have widespread problems of data sparse and polysemy,we propose a text sentiment classi-fication methods with a combination of latent semantic analysis(LSA)and support vector machine(SVM).The proposed method obtains the semantic information from the context of the words,thus eliminating the semantic bias of words which is caused by the existence of Polysemy in documentation set,improving the accuracy of the semantic representation of words and expanding coverage of words in the document.Experimental results show that the proposed method can effectively solve the probl-ems of existing sentiment classification approaches and improve the accuracy of sen-timent classification of text.2)To deal with the problem that existing tendency of view and prediction meth-ods ignore the impacts of the tendency of view on the popularity of a topic,this paper improves the existing methods by combining topic tendency analysis.Specifically,we present a hot topic discovery and prediction method that fuses the view of tendency factor.In topics discussion,the more diverse of the views or opinions are,the more intense the topics of discussion are.Thus the topic is more popular.Experimental results show that the results of the improved method get closer to the actual topic trends.