Research On Hot Topic Detection And Tracking In Internet Public Opinion

With the rapid development of Internet, it has become an important "place" forpeople to express their emotions and attitudes. Whether positive or negative networkpublic opinion has an important influence on social stability. Therefore, relevantdepartments of the State increase emphasis on the Internet public opinion, and theInternet public opinion monitoring system has come into being. It collects Internetinformation in real time, and intelligently analyzes the content of information. Themonitoring system provides good support for monitoring network public opinion,guiding the positive opinion and dealing with the crisis of public opinion to therelevant departments.We conduct our study on related technologies to build the network publicopinion monitoring system in this paper, such as natural language processing, datamining technology. At the same time, we have studied the existing public opinionmonitoring systems. On this basis, we constructed a public opinion monitoringsystem.Our study is focused on the key technologies of the Internet public opinionmonitoring system, including topic detection and tracking. These algorithms in thispaper are improved to adapt to reality monitoring environment. The contribution ofthis paper lies in the following three aspects:1. This paper presents an improved Single-Pass incremental clustering methodfor topic detection. A specific process of topic detection is studied in this paper, andthen we choose the text clustering as a key technology for its implementation. Weanalyze the principles and steps of several existing text clustering algorithms, andcompare the advantages and disadvantages of each algorithm. Finally, this paperselects Single-Pass algorithm for topic detection’s implementation. However, theSingle-Pass algorithm has some shortcomings, for example it’s sensitive to text inputsequence. On the basis of the analysis of the algorithm, it is improved on theaccuracy and response time, so it is able to adapt to large-scale text clustering.Finally, the comparative experiment proves that the proposed algorithm has a goodpractical application.2. An improved SVM algorithm applied to topic tracking is presented in thispaper. A specific process of topic tracking is studied in this paper, and then wechoose text classification as a key technology for its implementation. Through theanalysis of text classification methods and experimental comparison, we choose theSVM algorithm for its concrete realization. SVM algorithm is adapted to binaryclassification problems, so SVM is improved to solve multi-classification problem. Finally, the comparative experiment proves that the proposed algorithm has a goodpractical application.3. On the basis of the related technologies, we design the physical and logicalframework of Internet public opinion monitoring system. The detailedimplementation and function of system’s each module is described in this paper.Finally, we complete the system. The actual running of the system proves that theproposed design has a full-featured, fast, stable performance and so on. It provideseffective support for the monitoring and analysis of Internet public opinion.