Public Opinion Extraction Method Based On Text Emotional Computing Research

Along with the emergence of the Internet, a new way to express and spread the public opinion is given. More and more people express their opinions and views on some social issues by the internet. Because of the less restrictive, network forums is deeply loved by the people. But information on the network forum is complex and diverse, and part of information is bad. When the information is widely spread in the users, it is very easy to form public opinion and cause bad effect on people. If public opinion cannot be combed and untangled, it is easy to stimulate public sentiment and seriously endanger the social security and stability. Therefore, the government must master the leading power of public opinion to maintain the social security and stability.At present, the extraction of the public opinion is an effective ways to understand public feelings. Four elements are extracted, which are themes, holders, statement of holders and emotional tendencies of statement. According to the post features on the network forum, we concentrate on the extraction of emotional tendencies of statement, and analyze and extract the replies text emotion tendencies.The texts on the network forum often commence around many themes, the terminology is not standard and often contain network language. To solve the problem, a network polarity dictionary is expanded by adding a network terminology polarity dictionary and network colloquial polarity Dictionary in this thesis. On this basis, the judgment of sentence emotional tendencies is realized in full consideration of the negative words, adverbs, punctuation to the influence of sentence emotional tendentiousness.For text affective computing based on two themes (A,B), rules are set:if A is supported and B is denied, the polarity of the sentence is positive sign; if A is denied and B is supported, the polarity of the sentence is minus sign. The method in this paper can adjust overall tendency of the sentence by this rule, and realize affective computing for text and obtain emotional inclination of mass. Finally, this paper designs a system that realizes text orientation computing and displays public opinion. Through experiment, this method is proved to be effective.