Parvin Reisinezhad - Department of Computer Science and Engineering and IT Shiraz University, Shiraz, Iran
Mostafa Fakhrahmad - Department of Computer Science and Engineering and IT Shiraz University, Shiraz, Iran

Since the outbreak of the pandemic of coronavirus disease ۲۰۱۹ (COVID-۱۹), a great number of studiesin the form of questionnaires have been made available to a study population to evaluate the publicknowledge, attitudes, and practices (KAP). This research aims to apply the data from the social media andemploy the text-processing approach to achieve similar results, apart from the questionnaires. For thepublic knowledge and practice, we labeled around a thousand cases on a random basis, based onprinciples of medical guidelines of the World Health Organization where each tweet can meet zero orseveral items of the hygiene guidelines and employed Krippendorff’s Alpha Coefficient and achievedinter-annotator agreement of ۸۷%. For the public attitudes, we also similarly categorized ۱۰۰۰ instancesof viewpoints as positive, negative and neutral, where Krippendorff’s Alpha coefficient annotators’agreement of ۹۵% has been achieved between labelers at this point. We evaluated XLNet and BERT asour proposed models with several machine learning and deep learning approaches. Consequently, XLNetachieved to be of the highest efficiency. The use of social network data can reveal the truth in the sensethat people freely and with their knowledge express their opinions, which can solely explore the publicpoints of interest and reduce the negligence of responses.

Text mining, Sentiment analysis, Deep learning, KAP, Cognitive science