A Multilevel Graph Convolution Neural Network Model for Rumor Detection

Abstract

Rumor detection is a challenging task on social medias. When a post is propagated on social media, it usually contains four types of information: 1) content; 2) time of publishing; 3) structure of propagation; 4) social interaction. In most previous studies, the information has not been effectively combined to detect rumors. A multilevel graph convolution model including post level and event level is proposed to detect rumors in this paper. For post level graph convolution network based on propagation relationship, it uses a graph convolution network with rumor propagation graph to learn post level features. For event level graph convolution based on event interaction relationship, a graph convolution network with event relationship graph is applied to bridge post level features and event interaction information to obtain the feature representation of events. The experiment results shows that rumor detection accuracy of our model is 94.3%, which is superior to other newly models.