ARCNN framework for multimodal infodemic detection.

Area of Science:

• Artificial Intelligence
• Machine Learning
• Natural Language Processing

Background:

• Fake news and misinformation spread rapidly on social media, particularly during the COVID-19 pandemic.
• Automated detection technology is crucial for combating the harmful effects of online misinformation.
• Existing deep learning models require substantial datasets, which are scarce for coronavirus-related fake news.

Purpose of the Study:

• To develop and evaluate a novel framework for detecting fake news related to the COVID-19 pandemic.
• To address the scarcity of relevant datasets by creating the Coronavirus Infodemic Dataset (CovID).
• To leverage multimodal data (text and images) for enhanced fake news detection.

Main Methods:

• Development of the Allied Recurrent and Convolutional Neural Network (ARCNN) framework.
• Utilizing recurrent neural networks (RNNs) and convolutional neural networks (CNNs) for text and image analysis, respectively.
• Combining text and image modalities using early and late fusion techniques for prediction.

Main Results:

• Extensive research and performance evaluation of various RNN and CNN models on six coronavirus-specific fake news datasets.
• Demonstrated effectiveness of the ARCNN framework in detecting fake news through multimodal fusion.
• The proposed ARCNN models outperformed existing state-of-the-art fake news detection mechanisms.

Conclusions:

• The ARCNN framework provides a robust and effective solution for detecting COVID-19-related fake news.
• Multimodal analysis, combining text and image data, significantly improves fake news detection accuracy.
• The developed CovID dataset serves as a valuable resource for future research in combating infodemics.