Evolution Analysis Of Public Opinion Topics Based On ’Word-topic’ Coupling Network

Entering the web 2.0 era,the Internet has become an important medium for citizens to obtain news and express their opinions.In the network information dissemination,the rapid spread of news topics or emergencies poses a serious challenge to the government and other relevant functional departments.How to quickly track the news topics or the follow-up development of unexpected events is an urgent problem to be solved.This paper describes the evolution relationship between topics in the development process of public opinion events by observing the micro-characteristics of feature words in the evolution process,and then provides a theoretical basis for the relevant departments to guide and intervene in public opinion.This paper takes topic evolution as the main research object.Firstly,by observing the microcharacteristics of feature words in the process of drift,the paper analyzes the necessity of studying topic evolution from the perspective of words;A microblog data set is collected,the topic is automatically extracted through the topic model(LDA),and the mapping relationship between the word and the topic is captured to construct the ‘word-topic’ coupling network.Secondly,based on the‘word-topic’ coupling network,this paper analyzes the connotation of topic evolution and its index measurement from the three dimensions of topic intensity,topic status and topic drift path,including:the index measurement formula of topic intensity evolution considering the number of documents and the contribution of feature words to the topic;a two-dimensional status evolution index composed of weighted density index and weighted centrality index;and in order to detect the new topic which drifts from the original topic in the process of topic evolution,the formula of topic drift probability is proposed.Finally,the topic COVID-19 on the microblog platform from December 31,2019 to April 22,2020 is selected as an example to analyze the intensity evolution,status evolution and topic drift path.It is found that in the process of topic evolution,the change of feature words can reveal the evolution of topic from a micro perspective.The topic evolution analysis method based on ‘wordtopic’ coupling network can detect new topics in the process of public opinion topic evolution.