Graph Global Attention Network with Memory for Fake News Detection

Abstract: With the proliferation of social media, the detection of fake news has become a critical issue that poses a significant threat to society. The dissemination of fake information can lead to social harm and damage the credibility of information. To address this issue, deep learning has emerged as a promising approach, especially with the development of NLP. This study addresses the problem of detecting fake news on social media, which poses a significant challenge to society. This study proposes a new approach named GANM for fake news detection that employs NLP techniques to encode nodes for news context and user content and uses three graph convolutional networks to extract features and aggregate users' endogenous and exogenous information. The GANM employs a unique global attention mechanism with memory to learn the structural homogeneity of news dissemination networks. The approach achieves good results on a real dataset.