Propaganda Detection in Public Covid-19 Discussion on Social Media

As social media has grown exponentially during Covid-19, they have helped disseminate information, spread fake news and propaganda; thus provide a source of self-reported symptoms of illness (infected with Covid-19) in public discourse. This study presents a deep learning model tuned to RoBERTa and develop a precise model for detecting propaganda in text for multi-label, multi-class (MC-ML) classification in a specific domain/theme. Using data mining to covid-19 public discussion, we compare the models using long-short-term-memory (LSTM) and condition random field techniques with n-grams and TF-IDFs. Experimental results optimization improves modeling evaluation, and LSTM can accurately detect propaganda in public discussion. The MC-ML classification model has attained an accuracy of 82% with the proposed classifier, outperforming existing state-of-the-art techniques. Accordingly, this study assists IS researchers and practitioners in identifying and tracking propaganda on social media and provide furthering insight into data which is available to the research community for future research.