ANN: adversarial news net for robust fake news classification.

Area of Science:

• Natural Language Processing
• Artificial Intelligence
• Computational Social Science

Background:

• The proliferation of social media has accelerated the spread of fake news, posing significant risks to individuals and society.
• Effective fake news classification is crucial for mitigating the adverse impacts of misinformation.

Purpose of the Study:

• To develop an end-to-end framework for robust and resilient fake news detection.
• To enhance fake news classification performance by incorporating emoticon analysis and adversarial training.

Main Methods:

• Development of the Adversarial News Net (ANN) framework.
• Extraction and integration of emoticon meanings into the fake news detection model.
• Application of adversarial training to improve model robustness.

Main Results:

• The ANN framework demonstrated superior performance compared to baseline methods and previous studies.
• Adversarial training resulted in a 2.1% accuracy improvement over Random Forest and a 2.4% improvement over BERT.
• Emoticon analysis contributed to enhanced fake news classification accuracy.

Conclusions:

• The proposed ANN framework offers an effective solution for real-time fake news detection.
• Adversarial training and emoticon analysis are valuable techniques for improving the resilience of fake news detection models.
• The framework has the potential to significantly mitigate the societal harm caused by fake news.