Utilizing deep learning models for ternary classification in COVID-19 infodemic detection.

Area of Science:

• Computer Science
• Artificial Intelligence
• Public Health

Background:

• The COVID-19 pandemic was accompanied by a significant infodemic, characterized by the spread of misinformation and disinformation.
• Distinguishing between true, false, and uncertain information is crucial for public health responses.

Purpose of the Study:

• To evaluate the effectiveness of various deep learning models for ternary classification of COVID-19 infodemic content.
• To establish benchmark performance metrics for infodemic detection using machine learning.

Main Methods:

• Eight deep learning models were applied: fastText, recurrent neural networks (RNNs), convolutional neural networks (CNNs), and transformer-based models.
• Models were trained and evaluated on datasets to categorize records as true, false, or uncertain.

Main Results:

• Performance was assessed using precision, recall, F1-score, and overall accuracy.
• Confusion matrix analysis provided detailed insights into classification errors and model behavior.

Conclusions:

• Models with pretrained embeddings or simpler architectures generally outperformed complex models on the tested datasets.
• The findings suggest that efficient, simpler models are promising for COVID-19 infodemic detection, warranting further investigation.