ABSTRACT
In these modern computing days, the realm of sentiments analysis (SA) and opinion mining (OM) has captured enough attention due to the availability of massive though unstructured data available over miscellaneous social media platforms, blogs, e-commerce portals, and other similar digital resources. The outbreak of the COVID-19 pandemic in 2020 led to such excessive public discussions over various social media platforms, and people exhibited their emotions of fear, threat, hope, faith etc. in their posts. Twitter has been a very popular social media platform among the general public, celebrities, government personnel, entrepreneurs, etc. In the present study, people’s emotions and opinions about the current pandemic, expressed through Tweets, were extracted, interpreted, and analyzed through various machine learning (ML) techniques. We hypothesize a three-tier model for the whole research process. Being unstructured, uncertainty and imprecision are inherent in this data, and this needs to be normalized before extracting the vital information. Therefore, the Tweets dataset is passed through various data preprocessing treatments first for repairing and removing dirty data. The refined Twitter data are then visualized in a better way to seek interesting insights and trends. Next, five popular ML classifiers are employed for sentiments classification and comparative performance assessment through the Google Colab platform. Tweets are classified into negative, neutral, and positive genres based on polarity scores, and the ML classifiers are compared over the standard performance parameters like accuracy, precision, recall, and F-measure.
