An Analysis On Sarcasm Detection Over Twitter During COVID-19

The 2019 Coronavirus (COVID-19) significantly affected our society, the country, and the world. During the Corona time, people mostly spent their time on the internet and actively connected with other people through online social media (OSM) or game chatboxes. Due to extensive use of the internet and social media, the sharing of offensive content increases. OSM provides a platform where people freely express their opinion, emotions, and thoughts. Sometimes people share their feelings and thoughts sarcastically, wherein it signifies the opposite of what it states. Sarcastic content shared by people can vary in many forms, such as videos, images, podcasts, audio, and text. This research mainly focuses on Twitter text data extracted by the Twitter API during COVID-19 and investigates sarcastic content with negative sentiments during COVID-19. We extracted the data with some specific keywords like hashtag-related sarcastic information, sarcasm, irony, etc., and performed an offensive and aggressive nature analysis of people at this stage. We have used the linear support vector classifier (libSVM), Naïve Bayes, and Decision Tree for this analysis. The Decision Tree achieved the highest accuracy as compared to libSVM and Naïve Bayes. It can detect sarcastic content with up to 90% accuracy.