Chandra, Shantanu, et al. "Graph-based modeling of online communities for fake news detection." arXiv preprint arXiv:2008.06274 (2020).

This research introduces SAFER, a novel framework for detecting fake news on social media using graph neural networks (GNNs). SAFER is unique in its approach as it not only considers the content, structure, and dissemination patterns of online posts but also incorporates the properties of the online communities interacting with these posts. By aggregating information about the content, the sharing behaviors of users, and the users' social networks, SAFER provides a comprehensive analysis for fake news detection. The study also explores the use of relational and hyperbolic GNNs, which are new to the field of natural language processing (NLP) for community modeling, and demonstrates that SAFER outperforms existing text-based detection methods, achieving state-of-the-art results on fake news datasets from two different domains.