A Research Of Internet Public Opinion Monitoring-oriented Hot Topic Detection Technology

With the government and companies'increasing emphasis on internet public opinion monitoring, more and more public opinion monitoring system have been developed to assistant its user to deal with the public pressure or mass events on the Internet. Public opinion monitering system can gather related information real time, analyze the content intelligently, dig out the danger in public opinion timely, support the watching and disposing internet public opinion automaticly, as well as convenient its user and strongly help its user deal with public opinion correctly.In most of public opnnion monitering system, user gathers reports from internet sites periodically by customize source sites and system present the results to user after system'sanalysis and dispose. There are always lots of repetitious reports or reports describing the same event appearing between the results. This is not good for user either to look over, analyze and summarize the situation or to judge of the detail of one event and its heat. So it is necessary to improve the situation by using topic detection technology.In this paper we study the key technology of topic detection. We conclude that clustering is the most decisive step after learning the whole process of topic detection. We introduce the main principium of some clustering algorithms and analyze their advantages and disadvantages, then we pick single-pass clustering algorithm which is very popular in topic detection evaluating as our core algorithm. We study the difficulty and restrictive factors lying in single-pass algorithm's practicality, we set about a series of measures to improve its performance such as speed, accurateness and user intervention, and finally we obtain a well practical effect and fulfill the goal of topic detection.Aimed at the idiographic application, we study the hot evaluation and detection methods, and put forward integrated hot index based on focus degree and influence capability by meliorate index based focus degree. The integrated hot index can tally better with the goal of public opinion monitoring, with the integrated hot index the topic show a rank representing its importance laconically so that user can cope with the dynamic Internet better.We put forward the hot topic detection model which is very valuable to improve internet public opinion monitoring system by combining topic detection technology and hot evaluation method; at the same time, although research on topic detection and hot evaluation method is somekind of mature, we still face many difficulties in putting these technologies into practice, so it has very high of researching significance and value. At the end of this paper we proved the feasibility and validity of the hot topic detection model by running an instance and contrasts.