Rumor Detection by Analysing Tweets: A Comprehensive Framework

Online Social Network platforms are elaborately leveraged for the purpose of information or news collection during impactful events. Such information lacks credibility and proper verification which unravel rumors and can prove to be hazardous during high impact events. The lack of standardization of the process of restraint at the message level for online social network like Twitter, further proliferates the spread of rumor among huge number of users within a short span of time. In order to minimize the negative impact of rumor spread in Twitter during impactful events, this research has proposed a comprehensive rumor detection framework capable of detecting and classifying rumors from tweets generated during high impact events on Twitter. For every incoming tweet, attributes like content or text features of the tweet, user features- who tweeted and the contextual features of the tweet are considered for the comprehensive framework. All these different features are modelled by machine learning, deep learning, pre-trained word embedding and transformer model approaches. The experiments were performed on the four standard and benchmark rumor datasets- Boston Bombing, Pheme, Twitter15 & Twitter16, which are also used by other researchers for their state of the art research work. The results show an absolute maturation of neural networks considering both content and user features in detecting rumors in Twitter over other baseline models with respect to both accuracy and recall measures. This research has specifically brought out the importance of authenticity of users who are tweeting, in declaring a tweet as a rumor. Additionally, this research has implied that the popularity of the tweet and subjective keywords of a tweet also can facilitate the researchers in detecting rumors more accurately. The outperformance of XLNet as a transformer model in detecting rumors more accurately is another implication for the researchers.