Disinformation Detection Method Using Bidirectional GCN Based on Text and Emotion Embedding on Social Media

Key Points

• The study aims to develop a method for detecting unreliable information on social media using text and emotion embeddings.
• Utilized bidirectional GCN (Bi-GCN) for modeling propagation structures.
• Employed BERT for extracting text features from posts and comments.
• Applied BART to analyze the emotional distribution across threads.
• Integrated context, emotional distribution, and communication structure in the model.
• The proposed method outperformed existing disinformation detection techniques.
• Performance evaluations demonstrated significant improvements in accuracy and reliability of information verification.

Abstract