Fake News Detection Model Incorporating News Text and User Propagation

Fake news detection aims to detect the authenticity of news from different perspectives to maximize the performance of detecting fake news. In recent years, scholars have been engaged in the research of fake news detection, with their studies primary focusing on supervised learning. However, these studies require strenuous time and efforts to be spent on labeling datasets. To this end, this paper proposes a fake news detection model incorporating news text and user propagation named NT-UP by using unsupervised learning. NT-UP mainly consists of three components: text representation learning module, text feature extraction module and user propagation module. Specifically, the text representation learning module encodes news text and user tweets into a news vector and tweet vectors, respectively. The text feature extraction module is designed to extract news text features and in the user propagation module, the news propagation network is constructed by using news vector and tweet vectors, and then the graph feature is extracted from the news propagation network by using the graph contrastive learning based on unsupervised learning. The performance of the model is tested on real-world datasets, and experimental results show that the performance of NT-UP is better than that of other models.