MViR: Multi-View Visual-Semantic Representation for Fake News Detection

2026-02-26

AI Summary

Existing approaches to multimodal fake news detection struggle to adequately model the complex semantic relationships between images and text. To address this limitation, this work proposes MViR, a novel framework that introduces, for the first time, a multi-view vision-language modeling paradigm. MViR captures semantic variations of the same image under different textual contexts through a multi-view representation module, extracts multi-granularity image-text alignment features using pyramid dilated convolutions, and integrates diverse semantic cues via a multi-aggregation mechanism. Extensive experiments demonstrate that MViR significantly outperforms state-of-the-art methods across multiple benchmark datasets, confirming its effectiveness and superiority in multimodal fake news detection.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

With the rise of online social networks, detecting fake news accurately is essential for a healthy online environment. While existing methods have advanced multimodal fake news detection, they often neglect the multi-view visual-semantic aspects of news, such as different text perspectives of the same image. To address this, we propose a Multi-View Visual-Semantic Representation (MViR) framework. Our approach includes a Multi-View Representation module using pyramid dilated convolution to capture multi-view visual-semantic features, a Multi-View Feature Fusion module to integrate these features with text, and multiple aggregators to extract multi-view semantic cues for detection. Experiments on benchmark datasets demonstrate the superiority of MViR. The source code of FedCoop is available at https://github.com/FlowerinZDF/FakeNews-MVIR.

Problem

• fake news detection
• multi-view
• visual-semantic representation
• multimodal
• social media

Innovation

• Multi-View Representation
• Visual-Semantic Fusion
• Pyramid Dilated Convolution
• Fake News Detection
• Multimodal Learning