Co-Attention Based Multi-contextual Fake News Detection

Kapadia, Paritosh Saxena, Akrati Das, Bhaskarjyoti Pei, Yulong Pechenizkiy, MykolaIn recent years, the propagation of fake news on social media has emerged as a major challenge. Several approaches have been proposed to detect fake news on social media using the content of the microblogs and news-propagation network. In this work, we propose a method, named FND-NUP (Fake News Detection with News content, User profiles and Propagation networks), to detect fake news using users’ profile features, fake news content, and the propagation network. We use graph attention networks (GAT) to learn users representations using users’ profile features and news propagation networks. Next, we use co-attention technique to simultaneously learn the graph attention and the news content attention vectors, that will subsequently use to detect fake news. The derived co-attention weights allow our framework to provide the propagation graph-level and news article word-level explanations, respectively. We demonstrate that FND-NUP method outperforms state-of-the-art propagation-based and content-based fake news detection approaches.