Rumor detection with self-supervised learning on texts and social graph

Rumor detection has become an emerging and active research field in recent years. At the core is to model rumor characteristics inherent rich information, such as propagation patterns social network semantic post content, differentiate them from truth. However, existing works on fall short modeling heterogeneous either using one single information source only (e.g., network, or content) ignoring relations among multiple sources fusing content features via simple concatenation). Therefore, they possibly have drawbacks comprehensively understanding rumors, detecting accurately. In this work, we explore contrastive self-supervised learning sources, so reveal their characterize rumors better. Technically, supplement main supervised task of with auxiliary task, which enriches representations self-discrimination. Specifically, given two views a (i.e., encoding patterns), discrimination done by maximizing mutual between different same compared that other posts. We devise cluster-wise instance-wise approaches generate conduct discrimination, considering sources. term framework (SRD). Extensive experiments three real-world datasets validate effectiveness SRD for automatic media.