BIC: Twitter Bot Detection with Text-Graph Interaction and Semantic Consistency

Anonymous

Keywords: Twitter bot detection, text-graph interaction, semantic consistency

Abstract: Twitter bot detection is an important and meaningful task. Existing bot detection methods use either text modality to detect bots with anomalies in tweet patterns or graph modality to detect bots with abnormal clustering information. They do not allow text and graph modalities to interact with each other, which fails to learn the relative importance of the two modalities. As a result, these methods struggle to detect bots comprehensively. Besides, existing methods ignore the potential consistency within users' semantic information. In this paper, we propose a novel model named BIC that makes the text and graph modalities interactive. BIC also detects semantic consistency within tweet content. Specifically, BIC contains a text propagation module to learn text information, a graph propagation module to learn neighborhood information, and a text-graph interactive module to make the two interact. Besides, BIC contains a semantic consistency detection module to learn semantic consistency information from tweets. Extensive experiments demonstrate that our framework outperforms competitive baselines on a comprehensive Twitter bot benchmark. We also prove the effectiveness of the proposed interaction and semantic consistency detection.

Paper Type: long