Research And Design Of Public Opinion Warning System Based On Conceptual Graph

Since the beginning of the 21 Century, with the rapid development of the internet, the internet has become the main media of spreading public opinion information. The development of public opinion has the characteristics of massive, rapid and blind. So it is significant to use computer to analyze and monitor public opinion. In recent years, lots of research institutions and scholars have made a lot of research on public opinion analysis and made a lot of achievements. But because many reviewers use humorous discourse to comment something. Traditional methods of analyzing public opinion can only analyse the surface meaning of discourse and can’t dig out the deep meaning of comment. This situation can reduce the accuracy of public opinion analysis. Reference inference in pragmatic inference can effectively dig out the deep meaning of the discourse of the reviews. So this paper uses the relevance inference model to infer the deep meaning of the discourse for the accurate analysis of public opinion from the aspect of computer science. This paper mainly studies the following contents.(1)After studying a large number of reference inference theories, reference inference mode is summarized. And all kinds of knowledge expression methods are researched. The knowledge expression method of fuzzy conceptual graph not only has accurate and strict semantic definition and can express deep semantic knowledge, but also can carry out a variety of matching inference. The comment text of reviews and cognitive context knowledge are expressed by fuzzy conceptual graph which is suitable for reference inference mode. And the reference inference algorithm based on fuzzy conceptual graph matching is designed. This algorithm can automatically infer the discourse of the QA- model to deep meaning. Through the analysis of the experimental results, the algorithm has a better accuracy and smaller standard error.(2)Concept node in the traditional conceptual graph hasn’t shown tendency. Therefore, this paper researches the role of relation nodes in conceptual graph to sentence tendency. And traditional conceptual graph is improved by adding tendencious value and degree value to partial concept node in traditional concept graph. Tendentious conceptual graph is designed.(3)The traditional public opinion analysis mainly uses text clusteing method, classification method and sentence tendency analysis methods based on semantic. Text clusteing and classification methods can find hot topics, but it can’t analyze accurately tendency of comment text. Sentence tendency analysis methods based on semantic includes SBV polarity transfer algorithm and improved algorithm based on SBV. These algorithms can make up for text clustering and classification algorithm to a certain extent, but there are also many problems, which don’t consider the tendency of subject and verb-complement construction. Aimed at disadvantages of those algorithm, the algorithm of calculating attitude value based on tendentious conceptual graph is designed. A certain company’s comments text collected in internet are tested by the algorithm of calculating attitude value based on tendentious conceptual graph. Through the analysis of the experimental results, the algorithm has a better accuracy and smaller standard error.(4)On the basis of the above research, the prototype system of warning company public opinion based on the conceptual graph is designed. The prototype system mainly includes modules of data collection, preprocessing, reference inference, attitude value calculation and displaying and warning public opinion. At last, the prototype system is tested, and the system can complete the function of analyzing and warning company public opinion.The QA- mode’s comment text is inferred by the reference inference algorithm based on fuzzy conceptual graph matching to deep meaning. And the original comment text is replaced by deep meaning. Then, the attitude value of comment texts is calculated by the algorithm of calculating attitude value based on tendentious conceptual graph, so as to achieve the purpose of accurate analysis of public opinion.