Short Video is not only Video: Multimodal Unified Social Hypergraph Contrastive Enhancement for Fake News Video Detection
Abstract: Nowadays, fake short videos have seriously affected people's perception of news and situational awareness of event development. Previous work mainly focuses on the characteristics and dissemination of the news, and there is no in-depth mining of the social relationships and feature relationships of videos. This paper proposes a Multimodal Unified Social Hypergraph Contrastive Enhancement method MUHC for fake news videos detection. First, a unified social hypergraph is innovatively established for the representation of potential relationships in short videos. Meanwhile, a multimodal contrastive learning method for intra-modal and inter-modal relationships are designed to integrate different modalities. The above approach enhances data scalability while learning deeper about the potential relationships of the videos. Extensive experiments demonstrate that the method outperforms state-of-the-art on benchmark dataset.
Paper Type: Long
Research Area: NLP Applications
Research Area Keywords: multimodal applications; rumor/misinformation detection;
Contribution Types: NLP engineering experiment, Data analysis
Languages Studied: Chinese
Submission Number: 816
Loading