Trust-Aware Evidence Reasoning and Spatiotemporal Feature Aggregation for Explainable Fake News Detection

Fake news detection has become a significant topic based on the fast-spreading and detrimental effects of such news. Many methods deep neural networks learn clues from claim content message propagation structure or temporal information, which have been widely recognized. However, firstly, models ignore fact that information quality is uneven in propagation, makes semantic representations unreliable. Additionally, most do not fully leverage spatial structures combination. Finally, internal decision-making processes results are non-transparent unexplained. In this study, we developed trust-aware evidence reasoning spatiotemporal feature aggregation model for more interpretable accurate fake detection. Specifically, first designed module to calculate credibility posts random walk discover high-quality evidence. Next, perspective structure, an evidence-representation capture interactions granularly enhance reliable representation two-layer capsule network was aggregate implicit bias while capturing false portions source transparent manner. Extensive experiments two benchmark datasets indicate proposed can provide explanations results, also achieve better performance, boosting F1-score 3.5% average.