Deep Learning Driven System to Analyze Reliability of COVID Infodemic and News Articles

In the modern era, social media and electronic news play an important role in the dissemination of information across the world. The ease of sharing news articles and information over the electronic media has made it challenging to prevent the outspread of fake news articles. Thus, it has been observed that in the present situation of a pandemic the misleading news specifically about COVID-19 are increasing day by day. Multiple research groups identified this challenge and proposed machine learning-based binary classifiers for categorizing news articles into fake and true classes. But, none of them addressed the challenge of identifying the misleading news. Also, the existing research works do not focus on examining the reliability of the news about COVID-19. Moreover, there is a lack of complete system that provide an integration of front-end and back-end where the users can check the reliability of news article and get recommendation of the sources for validation of news articles. The authors in this manuscript propose the Deep Learning-based tool with multiclass classifier for classifying the news articles into fake, true and misleading. This tool is an integration of front-end and back-end that is equally effective in assessing the reliability of news about COVID-19 and other events. The classifier of the tool has been trained on the integrated dataset comprising general new articles from the globe as well as news articles about COVID-19. The trained classifier reported an accuracy of 88% in classifying the news into Fake, Misleading, and True classes. The higher accuracy than state-of-the-art models validate the acceptance of the tool for real-life applications.