Domain- and category-style clustering for general fake news detection via contrastive learning

Danke Wu, Zhenhua Tan, Haoran Zhao, Taotao Jiang, Ning Geng

Abstract: Highlights•Propose a domain- and category-style clustering framework for general fake news detection.•Capture the style features for fake news detection without using additional data preprocessing.•Develop multilevel contrastive loss to enhance discrimination ability and generalization ability.•Achieve state-of-the-art performance in both cross-event and cross-news-domain detection tasks.•Provide explainable results and style patterns for identifying fake news.