Research status of deep learning methods for rumor detection.

Area of Science:

• Social Media Analysis
• Artificial Intelligence
• Information Science

Background:

• Social media rumors pose significant societal harm.
• Deep learning methods are increasingly used for rumor detection in open networks.
• A comprehensive overview of current research is needed.

Purpose of the Study:

• To systematically review and categorize existing deep learning-based rumor detection research.
• To provide a multi-perspective analysis covering feature selection, model architecture, and research methodologies.
• To identify datasets, challenges, and future research directions in the field.

Main Methods:

• Feature selection methods are classified into content, social, and propagation structure features.
• Deep learning models are categorized as Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Graph Neural Networks (GNN), and Transformer-based models.
• Thirty works are synthesized into seven distinct rumor detection approaches, including propagation trees and adversarial learning.

Main Results:

• Analysis categorizes rumor detection techniques based on feature types and deep learning model architectures.
• A novel classification of seven rumor detection methods is presented, offering comparative insights.
• Available datasets are enumerated, facilitating further research and development.

Conclusions:

• This review provides a structured understanding of deep learning for rumor detection.
• It highlights diverse methodologies and offers a roadmap for future research in combating misinformation.
• The work aims to support researchers in advancing the field of social media rumor management.