Multi-party Collaborative Hate Speech Study on Social Media via Personalized Federated Learning

Junhao Yin, Xiran Hu, Geng Zhao, Guochang Wang

Abstract: Recently, amidst growing legislation focus and public sensitivity to data security, researchers and social media operators are keen to develop effective privacy-preserving and multi-party collaborative methods, to keep early intervention against various crisis. In this study, we propose a novel multi-party collaborative hate speech detection mechanism (MC-HSD) as a macro-guidance to address the challenge. Under MC-HSD, we propose a novel HateFL-pro framework based on federated learning. We demonstrate its effectiveness via extensive experiments.