Research Of The Hot Topic Dection Algorithm And Application Of Internet Public Opinion Analysis

Network public opinion is the attitude sets of cognitive emotional and behavioral tendencies which collected and transmissed from internet events that was stimulued by a variety other things.When an event transformed into network public opinion the first step was caused widespread concern of Internet user and became a hot topic. That is to say the study of hot topics detection is particularly importan in the domain of network public opinion analysis.The algorithm of hot topics detection in this thesis is mainly composed of two parts that is the topic detection and heat assessment. The topic detection algorithm is the improved classic incremental clustering algorithm of single-pass, the improved points include two aspects, the one hand is as text semantic correlation to be the standard of topic clustering, which is different from the traditional algorithm Clustered by text similarity. The algorithm in this thesis is further more consideration about the semantics interrelated to text which can improved the accuracy of topic detection. To calculates the correlation between the words with the text by selecting corpus to training PLSA model to obtain correlation matrix and then calculates the correlation between the text to text. The threshold selection method is the second improvement, as it because different characteristics due to the different topic categories corresponding to the topic of physical distribution. To different from the traditional threshold selection by selecting specific news in the field of pre-corpus to training PLSA model for the topic clustering threshold range in the field which can improve the field attribute. The heat assessment is calculated the concerntion of those topics.Finally, to designed and implemented a online hot topic detection system based on portal news site (Netease education news and Sina education news), which the purpose is to text the practical application on semantics related topics detection algorithm. The core module in this system include hot topic detection module, Web page data preprocessing module, information collection module and display module and other modules to achieve the application of improved topic online hot topics detection algorithm.