DAMMFND: Domain-Aware Multimodal Multi-view Fake News Detection

Abstract

Recently, multi-domain fake news detection has garnered increasing attention in academia. In particular, the integration of multimodal information into multi-domain fake news detection has emerged as a highly promising research direction. However, this field faces three main challenges: (1) Inaccurate domain identification, where predefined explicit identifiers fail to adapt to the inherent complexity of data; (2) Imbalanced multi-domain data distribution, which may induce negative transfer effects; and (3) Variable multi-domain modal contributions, indicating domain-specific differences in how various modalities influence news veracity assessments. To address these issues, we propose the Domain-Aware Multi-Modal Multi-View Fake News Detection (DAMMFND) framework. DAMMFND effectively extracts more accurate domain information through Domain Disentanglement, while simultaneously mitigating negative transfer between domains. Furthermore, DAMMFND introduces a Domain-Aware Multi-View Discriminator and a Domain-Enhanced Multi-view Decision Layer, which accurately quantify the contribution of domain information to multimodal, multi-view decision-making processes. Extensive experiments conducted on two real-world datasets demonstrate that the proposed model outperforms state-of-the-art baselines.