Multi-Context Based Neural Approach for COVID-19 Fake-News Detection

When the world is facing the COVID-19 pandemic, society is also fighting another battle to tackle misinformation. Due to the widespread effect of COVID 19 and increased usage of social media, fake news and rumors about COVID-19 are being spread rapidly. Identifying such misinformation is a challenging and active research problem. The lack of suitable datasets and external world knowledge contribute to the challenges associated with this task. In this paper, we propose MiCNA, a multi-context neural architecture to mitigate the problem of COVID-19 fake news detection. In the proposed model, we leverage the rich information of the three different pre-trained transformer-based models, i.e., BERT, BERTweet and COVID-Twitter-BERT to three different aspects of information (viz. general English language semantics, Tweet semantics, and information related to tweets on COVID 19) which together gives us a single multi-context representation. Our experiments provide evidence that the proposed model outperforms the existing baseline and the candidate models (i.e., three transformer architectures) and becomes a state-of-the-art model on the task of COVID-19 fake-news detection. We achieve new state-of-the-art performance on a benchmark COVID-19 fake-news dataset with 98.78% accuracy on the validation dataset and 98.69% accuracy on the test dataset.