Research On Topic Detection Of Microblog Public Opinion Based On Supernetwork Analysis

As a special kind of social media,microblog contains not only a large amount of text information,but also a variety of non-text information,such as microblog users,time series factors,emotional information,which together reflect the formation and evolution process of microblog public opinion topic.How to analyze the characteristics of public opinion elements in real-time from heterogeneous texts,and collaborate with topic model to study public opinion topic detection in micro-blog,so as to improve the efficiency of topic detection,and then to improve the level of public opinion supervision,has become the focus of current public opinion research.Based on the research of the mainstream topic detection analysis methods,this study believes that the main reasons for the current prolbems in the research of microblog public opinion topic detection,such as the model of microblog public opinion topic detection so simple,shortcomings revealed in the formation of themes,limited methods of topic detection,lack of analysis on distribution characteristics of viewpoints,lie in the lack of relevant research on the application of multi-source heterogeneous data in public opinion topic detection is based on this study.This paper proposes a research method of topic detection in microblog public opinion based on hypernetwork analysis.To solve the above problems,this paper conducts the following researches: Firstly,study the organizational method of multi-public opinion elements,draws lessons from traditional communication theory,takes into account the characteristics of micro-blog users(Who),content viewpoints(What),emotional characteristics(How)and time stages(When),constructs a micro-blog public opinion topic discovery supernetwork model,and explains the construction methods of sub-networks and the relationship between sub-networks.Secondly,research on microblog public opinion topic recognition based on supernetwork method.Using superedge of supernetwork analysis method,constructed a topic clustering method based on super-edge similarity.Based on subnetwork feature analysis method of supernetwork,constructed a multi-feature fusion clustering topic words extraction method.Thirdly,an topic evolution analysis based on supernetwork method is proposed.According to the characteristics of network structure of the supernetwork model of microblog public opinion topic detection,a method of constructing the index of topology analysis based on category is proposed,and then the dynamic evolution characteristics of public opinion topic are analyzed according to the time-varying characteristics of the index of topology.Lastly,it is empirical analysis.Based on the research of microblog public opinion topic detection,this paper takes the public opinion event of "seedless grape playing contraceptive drug" as an example,the reliability of this research method is verified and analyzed from two aspects of application domain analysis and method comparison analysis.The analysis results show that the results of application domain analysis of this method accord with the real situation of detection change under the public opinion event and the results are real and reliable.Moreover,this method have outstanding performance in multi-factor organizational ability,public opinion topic recognition ability and public opinion topic evolution analysis ability.This proves that the research model and analysis method proposed in this paper are scientific,reasonable and effective,and can be applied to the topic detection analysis of microblog public opinion with multi-factor characteristics.The innovation of this paper is to construct a microblog public opinion topic detedtion model based on supernetwork.Based on this model,a system of microblog public opinion topic identification and evolutionary analysis method is constructed by using supernetwork analysis method.