Misinformation detection based on news dispersion

Misinformation dispersion is of critical importance nowadays, especially in the media sector. With the increase of data sources and the rise of information dispersion within social media platforms, users face the dilemma of whether the information they come across is true or false. Technology has advanced significantly in recent years, aiming to detect misinformation by improving existing algorithms and introducing new features to achieve its early detection. In this paper, a Propagation, and opinion-related Feature-enhanced GNN-based framework (PFGNN) is presented. The aim of this framework is to capture all the correlations among the collected information and identify covered patterns in the collected data, in order to efficiently detect false and real information through the creation of an information propagation tree. This framework receives as input an enhanced propagation tree, including information collected from social media platforms, and news sources, along with propagation-related features extracted from the graph’s analysis, news representation, as well as social media users’ posts, in an attempt to capture user-specific behavior. This method has been evaluated on two data sets (i.e. PolitiFact and GossipCop), while has been compared with a baseline method using a form of user preferences information capture to tackle the misinformation detection challenge.