Monitoring And Early Warning Of Web Public Opinion Based On Semantic Mining

As the Internet has anonymity, fairness, freedom, spreading rapidly and many other advantages, the masses are often more willing to express their true thoughts on it. Internet as a virtual community, has become an important space where the public opinion generate and spread. Monitoring and controlling of the Internet public opinion has become an important means for the government to understand and divert the public opinion. It is difficult to achieve monitoring of the Internet with a manual way, because of the huge amount of information on the Internet.Thus a monitoring system on the base of computer network technology and automatic text analysis and processing technology is necessary. At present, there have been many researches on the monitoring system of the Internet public opinion.Existing monitoring system of Internet public opinion is based on the methods of statistics and key words. As ignoring the semantic information of the text, it often leads to inaccurate results. This article attempts to introduce the technology of Ontology and Semantic Computing into the domain of public opinion, in order to improve the performance of public opinion monitoring systems.At first,it introduces the technologies of Web information gathering and pretreatment, including an improved method of text extraction from Web pages. Then the technologies of text mining, Ontology and the Semantic Web are introduced. With the help of domain experts and available reference, it builds a domain ontology of Internet public opinion, and uses this domain ontology to implement the technology of query expansion.Then it applies this technology to the applications of Internet public opinion retrieval. In this paper,it proposes an improved text similarity measure method based on HowNet, and an improved similarity-based text clustering algorithm. Finally, based on the existing research it establishes a index system to evaluate internet public opinion, and uses a kinds of generalized linear regression model with convex constraint and unknown dependent variable to implement mathematical modeling for the Internet public opinion evaluation. And through comparing its assessment findings with the results published by the "yq.people.com.cn", demonstrates its technical feasibility to a certain extent.