Research on Online Rumor Detection Based on Tweet Propagation Patterns and Cross-Modal Features
[Research purpose]To effectively manage online rumors and reduce the threat of online rumors to social stability,we propose to fully integrate the multimodal information of tweets and the propagation pattern information to accurately identify rumors.[Research method]We propose a rumor detection model(PPCMRD)that integrates tweet propagation pattern information and cross-modal fea-tures.In terms of tweet propagation feature mining,the first step is to complement the tweet propagation graph by inferring potential con-nections,followed by encoding multiple propagation patterns of tweets using the bidirectional signed graph attention module,and then cap-turing the complementary information between pattern features through the propagation pattern information fusion module to obtain the propagation features of the tweet.In terms of integrating the multimodal features,this model integrates the text,image,and tweet propa-gation features of the tweet,and employs the cross-modal co-attention mechanism to capture the complementary relationship between dif-ferent modal information and get the final embedding representation of the tweet to determine whether it is a rumor or not.[Research con-clusion]The experimental results on two public datasets demonstrate that the proposed approach could effectively detect rumors and outper-forms the current baseline models.
online rumorrumor detectiononline rumor detection modeltweet propagation patternscross-modal features fusion
万方数据
维普
