Scholarlink

<p>The proliferation of online social networks has become a major conduit for fake news dissemination, eroding public trust and challenging information ecosystems. Existing fake news detection (FND) methods fail to adequately leverage latent complementary relationships between multimodal features, leading to suboptimal information fusion. This paper proposes a novel method for multimodal fake news detection, named complementary perspectives enhancement via hierarchical graph network (CPEHGN). Specifically, we design three hierarchical graphs to explicitly establish both intra-modal correlations and cross-modal associations within news clusters. The features from different modalities are further updated with modality-specific graph convolution networks (GCNs), which enhance the complementary relationships from both visual and textual perspectives. Moreover, a self-attention module is incorporated in the textual modality encoding phase to capture hierarchical semantic information. Extensive experiments on two benchmark datasets validate the superiority of our approach with the accuracy of 94.9% and 89.6%. Notably, CPEHGN is designed to leverage high-performance computing infrastructures and parallel processing environments, endowing it with the potential to handle the large-scale news data commonly encountered in real-world scenarios. The source code is available at <a href="https://github.com/wenbin-zheng/CPEHGN">https://github.com/wenbin-zheng/CPEHGN</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Complementary perspectives enhancement via hierarchical graph networks for multimodal fake news detection

  • 期刊论文

  • 发表日期：

    2026年1月6日

查看全文
  • Zhijian Huang,
  • Wenbin Zheng

摘要

The proliferation of online social networks has become a major conduit for fake news dissemination, eroding public trust and challenging information ecosystems. Existing fake news detection (FND) methods fail to adequately leverage latent complementary relationships between multimodal features, leading to suboptimal information fusion. This paper proposes a novel method for multimodal fake news detection, named complementary perspectives enhancement via hierarchical graph network (CPEHGN). Specifically, we design three hierarchical graphs to explicitly establish both intra-modal correlations and cross-modal associations within news clusters. The features from different modalities are further updated with modality-specific graph convolution networks (GCNs), which enhance the complementary relationships from both visual and textual perspectives. Moreover, a self-attention module is incorporated in the textual modality encoding phase to capture hierarchical semantic information. Extensive experiments on two benchmark datasets validate the superiority of our approach with the accuracy of 94.9% and 89.6%. Notably, CPEHGN is designed to leverage high-performance computing infrastructures and parallel processing environments, endowing it with the potential to handle the large-scale news data commonly encountered in real-world scenarios. The source code is available at https://github.com/wenbin-zheng/CPEHGN.