Sentiment Analysis of COVID-19 Tweets Through Flair PyTorch, Emojis, and TextBlob

Abstract

In the current decade, the economy and health have been significantly impacted globally by the pandemic disease named Coronavirus Disease 2019 (COVID-19). People need to stay indoors at this time, which causes them to grow more dependent on social media and use these online channels to communicate their feelings and sympathies. Twitter is one of the familiar social media and micro-blogging platforms in which people post tweets, retweet tweets, and communicate regularly, offering an immense amount of data. Popular social media have evolved into an abundant information source for sentiment analysis (SA) on COVID-19-related issues. Hence, SA is used to predict the public opinion polarity that underlies various factors from Twitter during lockdown phases. Natural language processing (NLP) has been utilised in this study to manage the SA and employ specific tools to codify human language and its means of transmitting information to beneficial findings. This proposed method for Twitter SA is concentrated on all aspects by considering the emoji provided and leveraging the Flair Pytorch (FP) technology. Since extracting emojis and text is implanted with sentiment awareness, it surpasses cutting-edge algorithms. In this research, the ‘en-sentiment' module is introduced in the FP method for tokenisation and text classification that assists in diverging the sentence with respect to words, namely positive or negative as sentiment status for the tweets. Thus, it is evaluated by the confidence score of the FP method and compared with the existing textblob method.