Research On The Key Technologies In Topic Detection And Tracking Of Internet And Sentiment Analysis

With the highly popularity of the Internet, the impact of the Internet on our real lifeis also growing. In the face of the growing mass information in the Internet, when userswant to timely track a theme, the existing Internet search engine is insufficient to meetthe needs of users, the system appears can be very good to help users detect, track andanalysis on the theme of public opinion monitoring analysis.This paper mainly analyzed the technology of network public opinions’ topicdetection, tracking and sentiment analysis. First, this paper studies the research status ofthese several technology. Then this paper carefully studied some important ones ofseveral technologies, such as, text representation model, Chinese analysis, feature weightcalculation methods and text classification. On this basis, the innovation of this paper asfollows.First, the paper improved vector space model and text classification algorithms byconsidering the evolutionary relationships between seed events and novel events. At thesame time, these improvements are also based on the model of the topic detection andtopic tracking. The model and algorithms are mainly for the phenomenon of theme driftin the monitoring analysis system of public opinion. The algorithm proved to be couldsolve the questions of low accuracy of topic detection and tracking by public drift in acertain extent by experimental analysis.Secondly, the paper analyze based on the emotion tendency of event multi-faceted.Currently, the majority of sentiment analysis module in the monitoring and analysissystem of public opinion is designed to get emotional polarity of the overall evaluationinformation for a particular topic or things, while ignoring the multi-faceted nature of thething itself. This paper improved emotional tendentious analysis modules’ function bytaking emotional statement emotional theme or things evaluation information of localemotion polarity.Finally, this paper designed a monitoring system of public opinion on the Hadoopplatform. The system is mainly for the current flood of Internet information, data storage and computing become an important part of the system of monitoring public opinionanalysis.This paper mainly studied and explored the technology of topic detection andtracking, emotional tendentious analysis. There have in-depth understanding of the twotechnologies. This paper put forward improved models and algorithms based on theexisting questions. It provides a certain amount of technical support for monitoringpublic opinion analysis research and has important theoretical significance.