Topic Detecting And Tracking Research Of Internet Public Opinion

In recent years,Internet public opinion has attracted much attention in China,and the frequency of Internet public opinion events has an increasing impact on the order of political life and social stability.In this situation,the relevant departments of the government and enterprises are paying more and more attention to the management of Internet public opinion,and they are required to strengthen monitoring efforts and pay close attention to developments.News topic tracking is an important part of Internet public opinion monitoring.The common solution to this task is topic detection and tracking(TDT),but the existing TDT method has great limitations due to the following reasons.One is that the object of Internet public opinion news topic tracking is a specific topic,and a human intervention mechanism based on keywords needs to be provided,while The traditional TDT method to track all topics is an unsupervised process.The other is that news topic tracking of Internet public opinion needs to sort out the development context of topics,but most of the traditional TDT methods have not studied this.Therefore,this paper mainly studies the topic detection and tracking method for Internet public opinion,and the main work is as follows:(1)A topic tracking method based on keywords is proposed.This method addresses two difficulties in topic tracking: low-frequency but high-importance words cause poor tracking effects and fixed keywords cannot cope with dynamic topic tracking,use topic-oriented keywords to represent news text to improve topic tracking effect,and use word-based vitality update strategies to dynamically adjust topickeywords.By conducting relevant experiments on the three hot topics of Internet public opinion recently,it is shown that this method can effectively track the dynamics of Internet public opinion topics.(2)On the basis of the previous method,an overall design of the topic detection and tracking method for Internet public opinion was designed,and a topic development context construction method based on heat calculation was proposed.This method detects topic events of appropriate granularity from topic news to construct topic development context.In view of the problem that the topic development context constructed by the current research methods is not intuitive and logically incoherent,this method uses a topic tree with a backbone and branch structure to present the topic development context from the perspective of the development rule of the Internet public opinion topic,and through the heat calculation to identify key event nodes of the topic.Experiments show that the topic development context constructed by this method performs well and has a clear structure,which can help users intuitively understand the topic development process.