A Rumor Detection Method Integrating BERT and Topic Model
[Purpose/significance]In the process of rumor detection,the contextual semantic features and topic semantic features of the text have not been sufficiently addressed.This study proposes a rumor detection method that combines BERT and topic models to en-hance the effectiveness of rumor detection.[Method/process]Utilizing the BERT model to extract dynamic contextual semantic fea-tures from text,and employing topic modeling techniques to mine topic-based semantic features,this paper also integrates both Weibo influence and user credibility features.In the design phase of the Weibo influence feature,time-sensitive factors are taken into consid-eration.The aforementioned features are fully integrated to build a rumor detection model.[Result/conclusion]An empirical analysis was conducted using real-world data from Weibo,and the experimental results demonstrate that the proposed method exhibits superior performance in rumor detection tasks,achieving a peak accuracy of 93.68％.This represents improvements of approximately 9.92％,3.29％,and 7.75％compared to the best-performing traditional machine learning methods,deep learning methods,and feature fusion methods,respectively,enabling effective rumor detection.Furthermore,considering the timeliness factor during the design phase of Weibo influence features contributes to enhancing the efficacy of rumor detection.[innovation/limitation]There is still room for im-provement in rumor detection performance.In the future,it may be worthwhile to explore incorporating user comment content features to further enhance the effectiveness of rumor detection.
NETL
NSTL
万方数据
