Deep Learning Approach to COVID-19 Misinformation on Social Media with Multimodal and Multilingual Deep COVID-Fake Sense Model

The dissemination of false information on social media platforms has posed a serious threat to public health initiatives and the welfare of society during the COVID-19 pandemic. In response, this work introduces DeepCOVID-FakeSense, an innovative multimodal sentiment analysis method for identifying COVID-19 disinformation. DeepCOVID-FakeSense combines a Graph Neural Network (GNN) model for picture sentiment analysis and a BERT + Bi-LSTM model for text sentiment analysis using cutting-edge deep learning techniques. Textual analysis in the study makes use of the COVID19 misinformation Dataset, and image sentiment analysis makes use of the MM-COVID: A Multilingual and Multimodal Data Repository for Combating COVID- 19 Disinformation dataset. Through the analysis of textual and visual content from these datasets, DeepCOVID-FakeSense offers a comprehensive comprehension of the sentiments present in social media posts related to COVID-19. The architecture of the model smoothly combines textual and visual elements, enhancing its capacity to identify minute details and contextual cues that point to false information. With the ability to categorise content as either real or fake news, DeepCOVID-FakeSense is an effective tool for combating COVID-19 disinformation. Experimental assessments validate the effectiveness of DeepCOVID-FakeSense in identifying misinformation with exceptional precision 82.95and recall 82.90%, achieving an overall accuracy of 82.93% and an F1 score of 82.92%. By illuminating the sentiment dynamics of COVID-19-related content, DeepCOVID-FakeSense empowers stakeholders to combat misinformation, protect public health, and promote informed decision-making in the digital environment.