Attention-Based Deep Learning Models for Detecting Misinformation of Long-Term Effects of COVID-19

During the COVID-19 pandemic, the surge of misinformation on social media threatens public understanding and epidemic prevention policies. Even as the pandemic is being controlled, long-term COVID-19 and reinfection risks still need to be included in COVID-19 policies and information. This study presented a deep learning approach to detect fake news related to the long-term influences of COVID-19. The data is collected and refined from reliable open sources with data processing techniques. Then, the various attention-based deep learning models like HAN, BERT, and XLNet are trained to detect misinformation about the long-term effects of COVID-19 based on the collected data. The F1 score reached 94.96%, showing the strong performance of the deep learning models. The method demonstrated high effectiveness in identifying such false content, contributing automatic tools for detecting misinformation on the long-term impacts of the COVID-19 pandemic.