Hot Topic Discovery And Polarity Analysis For Network News

With the rapid development of Internet media and the enhancement of the Internet penetration,there are a large number of news reports on the Internet.Reports on the same news topic are often scattered displaying in the major news sites.And not all news topics will be concerned.In addition,as the Internet more emphasis on interactivity,people tend to express their opinions and attitudes on the Internet.Facing with the massive news reports and reviews on the Internet,It is very important to classify numerous and scattered reports as various topics,display all the news on the same topic in a more centralized way,and then screen out hot topics and analyze their polarity.In this thesis,the real news data is the research object of hot topics discovery and polarity analysis.In order to detect current hot topics from the massive news reports,an algorithm for discovering hot topics is proposed in this thesis.As text clustering is a key step in topic discovery,this thesis focuses on improving text clustering algorithm and improves the accuracy of topic discovery by batch,multi-vector and quadratic clustering.After obtaining each topic,the heat value of each topic is calculated by the method of evaluating the topic heat proposed in this thesis,and each topic will be ordered according to the heat value.So that hot topics are screened out.On the basis of detecting hot topics,in order to get people's views and attitudes towards hot topics,this thesis presents a method of polarity analysis based on polar dictionary and semantic rules.This thesis builds a polar dictionary based on the open-source emotional dictionary.In the process of building the polar dictionary,it considers the influence of modifiers,network terms and evaluation objects on the results of analyzing.In addition,this thesis summarizes semantic rules used in polarity analysis.After that,a method of extracting features based on polar dictionary and semantic rules is used to extract emotion features in comments.Then,calculating the polarity of comments according to the method of calculating polarity proposed in this thesis,so as to obtain the polarity of each hot topic.Finally,the algorithm for discovering hot topics and the method of polarity analysis are verified by two applications.The experimental results show that the improved algorithms are feasible and effective.