Research on Opinion Leader Identification Model Based on Text Deep Clustering
The crime caused by online public opinion events is on the rise.However,traditional opinion leader identification methods are usually based on metadata such as user information,forwarding,and comment count,ignoring key information such as network structure and text content,and lacking opinion leader opinions,which can easily lead to biased results.To address the above issues,a semantic clustering-based opinion leader recognition model is proposed.This model can cluster user texts by the BERT-LDA&DEC algorithms,group opinion leaders according to different sub topics,and extract keywords.An indicator system for grouped users is established from three aspects:network topology,personal attributes,and activity level.The entropy weight gray correlation method is used to evaluate user indicators.Finally,a comprehensive analysis is conducted based on clustering keywords.Experiments have shown that this method can effectively identify opinion leaders and their viewpoints in different subtopics of microblog topics.
Deep Embedded Clustering(DEC)opinion leadersentropy weight grey correlation methodBERT-LDA
NETL
NSTL
万方数据
