Rumor detection using BERT-based social circle and interaction network model

Social media rumors, which spread uncontrollably, are the world's big problem and therefore there is a great need for effective early detection tools. However, despite the widely known influence of social circles on information disemmination, most existing detection models overlook it. This study addresses this gap by examining three key components such as: BERT for Tweet Analysis for Semantic Understanding, Social Circle and Interaction Tree Embeddings for Network Context. The first step in doing so is to exploit BERT in order to obtain subtle semantic meanings from tweets that could provide insights into rumor structures and their content. Second, social circle data capture the user's circle as well as their involvement with community towards leading rumors diffusion. Finally, interaction embeddings capture network dynamics thus providing useful contextual information to understand how rumor propagates over network paths. These components are integrated to develop a comprehensive algorithm for rumor detection. Our algorithm combines tweet's semantic analysis using BERT and provides network context from social circle and interaction embeddings. Experiments on real datasets demonstrate that our approach outperforms all others in terms of accuracy and efficiency thereby showing its effectiveness over existing methods. The summary of this research implies that BERT, social circles, and interaction patterns are substantial in a comprehensive rumor detection framework for online platforms. Adopting this method will facilitate us to detect rumors from their inception, thereby prevention is taken as one measure against the information mismanagement in social media networks.