Interpretable Fake News Detection on Social Media

Abstract

With the development of information technology, public opinion can quickly spread to all over the world, permeate every corner of social life, and have a great impact on human's lives. Extracted from large-scale and multi-mode social media, user-generated information is anonymous and noisy. It is found that users' social interaction helps to detect fake news. At present, most methods focus on effectively detecting fake news with potential characteristics, but most models are lack of interpretability. Therefore, this paper studies the interpretable detection of fake information in public social media platform, and proposes a model based on in-depth sentence-comment interactive reasoning network. The model uses fake information content and user comments to capture the most worth checking information sentences from user comments, in order to detect fake information and provide some explanation. This paper solves the following challenges: (1) how to detect fake news while improving the detection performance and interpretability; (2) how to extract the correlation between false content and user comments in the social media platform.