Scholarlink

<p>The rapid development of social media has led to the unguaranteed authenticity of news, and the role of fake news detection in cybersecurity governance has become increasingly prominent. In this paper, we mine information from multiple modalities, such as the text and images of news, and propose a multi-modal fake news detection model based on the multi-stage domain adaptation for the differences existing between source and task domains and between different modalities. The multi-modal feature extraction network of BERT combined with EfficientNet is used to deeply analyze the features of social media data, and the multi-modal domain adaptation network is used to reduce the domain shift of different domains and different modalities of news data and to capture the correlation between events by adversarial ideas. Experimental results on public datasets of Weibo and Twitter show that the model significantly improves the effectiveness of the fake news detection task.</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

Fake news detection based on multi-modal domain adaptation

  • 期刊论文

  • 发表日期：

    2025年1月4日

查看全文
  • Xiaopei Wang,
  • Jiana Meng,
  • Di Zhao,
  • Xuan Meng,
  • Hewen Sun

摘要

The rapid development of social media has led to the unguaranteed authenticity of news, and the role of fake news detection in cybersecurity governance has become increasingly prominent. In this paper, we mine information from multiple modalities, such as the text and images of news, and propose a multi-modal fake news detection model based on the multi-stage domain adaptation for the differences existing between source and task domains and between different modalities. The multi-modal feature extraction network of BERT combined with EfficientNet is used to deeply analyze the features of social media data, and the multi-modal domain adaptation network is used to reduce the domain shift of different domains and different modalities of news data and to capture the correlation between events by adversarial ideas. Experimental results on public datasets of Weibo and Twitter show that the model significantly improves the effectiveness of the fake news detection task.