Rumor detection in social network based on user, content and lexical features

Area of Science:

• Computer Science
• Social Network Analysis
• Artificial Intelligence

Background:

• Social networks enable rapid news diffusion, making data verification difficult.
• Unverified information, such as fake news and rumors, can cause significant harm to individuals and organizations.
• Existing rumor detection methods primarily focus on temporal dynamics, yielding moderate accuracy.

Purpose of the Study:

• To improve the accuracy of rumor detection on social media platforms.
• To explore the effectiveness of post-wise features, including user-based, content-based, and lexical-based features, alongside post sequences.
• To propose a novel deep learning framework for combating rumor diffusion.

Main Methods:

• Utilized word embedding with bidirectional long short-term memory (BiLSTM) for sequential data processing.
• Combined BiLSTM outputs with post-wise features using a multilayer perceptron (MLP).
• Developed and tested a framework integrating multiple feature types and deep learning models.

Main Results:

• The proposed framework demonstrated a notable improvement in accuracy compared to existing state-of-the-art approaches.
• The integration of diverse features and deep learning models proved effective in enhancing rumor detection.
• Experiments on a real-world Twitter dataset validated the framework's performance.

Conclusions:

• The developed framework offers a more accurate approach to rumor detection on social media.
• Incorporating post-wise features alongside advanced deep learning techniques is key to mitigating fake news.
• This research contributes to combating the adverse effects of rumor diffusion in online environments.