Scholarlink

<p>Current multimodal fake news detection methods usually extract and integrate news text and image features for detection, lacking comprehensive data integration. In addition, the large amount of noise generated by multi-source data also brings additional challenges to fake news detection. To address these issues, we propose a novel multi-dimensional feature collaborative fusion network, named MFCF. This model first constructs two branch networks to learn news content knowledge, including image frequency and spatial domain features, text semantics and sentiment features. A cross-modal co-attention fusion module is designed, which effectively eliminates noise in multi-source data through dynamic and context-aware feature fusion strategies, thereby extracting key information for fake news detection. Furthermore, we designed a cross-modal consistency learning branch network based on contrast language-image pre-training to enhance the model’s ability to detect image-text consistency. MFCF outperforms existing technologies on Weibo, Twitter and GossipCop datasets, achieving higher accuracy rates and surpassing the baseline.</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

Multi-dimensional feature collaborative fusion networks for multimodal fake news detection

  • 期刊论文

  • 发表日期：

    2025年5月26日

查看全文
  • Lixia Ji,
  • Jingmei Feng,
  • Shijie Xiao,
  • Han Zhang

摘要

Current multimodal fake news detection methods usually extract and integrate news text and image features for detection, lacking comprehensive data integration. In addition, the large amount of noise generated by multi-source data also brings additional challenges to fake news detection. To address these issues, we propose a novel multi-dimensional feature collaborative fusion network, named MFCF. This model first constructs two branch networks to learn news content knowledge, including image frequency and spatial domain features, text semantics and sentiment features. A cross-modal co-attention fusion module is designed, which effectively eliminates noise in multi-source data through dynamic and context-aware feature fusion strategies, thereby extracting key information for fake news detection. Furthermore, we designed a cross-modal consistency learning branch network based on contrast language-image pre-training to enhance the model’s ability to detect image-text consistency. MFCF outperforms existing technologies on Weibo, Twitter and GossipCop datasets, achieving higher accuracy rates and surpassing the baseline.