Joint Stance and Rumor Detection in Hierarchical Heterogeneous Graph.

Area of Science:

• Computer Science
• Social Media Analysis
• Artificial Intelligence

Background:

• Emerging social media platforms frequently host large volumes of unverified information, including fake news and rumors, which can have significant negative consequences.
• Existing rumor detection methods often overlook the deep correlation between stance distribution and rumor veracity, treating these tasks separately or using basic multitask learning.
• Current approaches heavily depend on handcrafted features and extensive labeled data, hindering early and few-shot rumor detection.

Purpose of the Study:

• To develop a novel framework that jointly addresses stance and rumor detection for improved accuracy in identifying unverified information.
• To explore the profound correlation between stance distribution and rumor veracity beyond conventional multitask learning.
• To reduce the reliance on handcrafted features and large labeled datasets for effective rumor detection.

Main Methods:

• Constructed a hierarchical heterogeneous graph by linking posts with shared high-frequency words to enable cross-topic feature propagation.
• Formulated stance and rumor detection as multistage classification tasks within a unified framework.
• Proposed a multigraph neural network (MGNN) to jointly update node embeddings driven by both stance and rumor detection, capturing rich attribute and structural information.

Main Results:

• The proposed multigraph neural network framework significantly outperformed state-of-the-art methods on both stance and rumor detection tasks using real-world Twitter and Reddit datasets.
• Experimental results demonstrated the method's superior performance even with limited labeled data, indicating effectiveness in few-shot and early detection scenarios.
• The approach showed improved interpretability compared to existing methods.

Conclusions:

• The developed multigraph neural network effectively captures the intricate relationship between post stance and rumor veracity for enhanced detection.
• This method offers a more robust and data-efficient solution for combating the spread of misinformation on social media platforms.
• The framework provides better interpretability and requires less labeled data, making it practical for real-world applications.