Birds of a Feather Rumor Together? Exploring Homogeneity and Conversation Structure in Social Media for Rumor Detection

Abstract

Rumors in social media represent a severe problem prevailing in today's society. Previous studies on automated rumor detection have shown that the topological information specific to social media is a vital clue for debunking rumors. However, existing automatic rumor detection approaches either oversimplify the graph structure or ignore this crucial clue. To address this issue, we propose a model that explores homogeneity and conversation structure to identify rumors. Our model learns more comprehensive and precise representations by modeling follower-following relationships of users, simulating the propagation layout of tweets, and connecting responders' behavior. The experimental results on two public Twitter datasets show that our model's performance outperforms other state-of-the-art baseline models. Furthermore, the experimental results prove our hypothesis that birds of a feather rumor together. The results demonstrate that both the conversation structure and the friend network's homogeneity are significant for checking the veracity of a suspicious tweet.

MVAN: Multi-View Attention Networks for Fake News Detection on Social Media

“…[27] built a conversation structure from source tweet and user comments, and used GNN to encode it. Li et al (2020) [28] crawled user-follower information and built a friendly network based on the followfollowers relationship. We compare our work and the most relevant studies in Table 1.…”

Rumor Detection Based on Attention CNN and Time Series of Context Information

“…The number of input parameters is large and this model occupied a large number of resources. Some scholars have also improved the LSTM model [24,25] by combining word attention and context information, or mining the homogeneity and dialogue structure features of rumors which has achieved good results. However, the relationship between the different life cycles and the life cycle of the development of microblog events has not been considered.…”

Catch me if you can: A participant-level rumor detection framework via fine-grained user representation learning

“…For instance, Bian et al proposed a GCN-based model to learn global structural relationships of rumor dispersion (Bian et al, 2020). Li et al also employed a graph-based model to explore the homogeneity and conversation structure in rumor spreading (Li et al, 2020). Yang et al proposed an adversarial learning-based model, taking into account the camouflages of rumors from an adversarial perspective (Yang et al, 2020).…”