Neighbours and Kinsmen: Hateful Users Detection with Graph Neural Network

With a massive rise of user-generated web content on social media, the amount of hate speech is also increasing. Countering online hate speech is a critical yet challenging task. Previous research has primarily focused on hateful content detection. In this study, we shift the attention from hateful content detection towards hateful users detection. Note, hateful users detection can benefit from users’ tweets, profiles, social relationships, but the real benefit is that it can be aided by Graph Neural Networks (GNN). Typical Graph Neural Networks, such as GraphSAGE, only considers local neighbourhood information and samples the neighbourhood uniformly, thus they lack the ability to capture long-range relationships or to differentiate neighbours of a node. In this paper, we present HateGNN – a GNN-based method to address these two limitations. Our proposed method relies on the notion of latent neighbourhood, as well as systematic sampling of the neighbourhood nodes. The experimental results demonstrate that HateGNN outperforms state-of-the-art baselines in the task of detecting hateful users. We also provide a detailed analysis to demonstrate the efficacy of the proposed method.