关键词: 谣言检测, 用户属性, 图注意网络, 多模态门控网络, 传播网络

Abstract: Abstract: In view of the conditional constraints such as short content of messages on social media platforms, the large amount of empty forwarding in the transmission structure, and the mismatch between user roles and content. this paper proposed a rumor detection model GMB_GMU based on user attribute information and message content in the propagation network. In the model, user attributes are used as nodes and propagation chain as edges to construct user propagation network. Graph attention network (GAT) is introduced to obtain enhanced representation of user attributes. At the same time, based on the user propagation network, the Node2 Vector is used to get the user structure representation, and the mutual attention mechanism is used to enhance it. In addition, BERT is introduced to establish the content representation of the source post. Finally, the multi-model gating network GMU is used to integrate the user attribute representation, structure representation and source post representation to obtain the final message representation. Experimental results show that the model’s accuracy reached 0.952 in the public dataset named Weibo and can effectively identify rumor events, the effect of this model is significantly better than the propagation algorithm based on recurrent neural network(RNN) and other neural network benchmark models.

Key words: rumor detection, user attribute, graph attention network, gated ,multimodal network, propagation network