Automated Classification of Societal Sentiments on Twitter With Machine Learning
Piyush Vyas, Martin Reisslein, Bhaskar Prasad Rimal, Gitika Vyas, Ganga Prasad Basyal, Prathamesh Muzumdar
Abstract
The rapid growth in information sharing on social media has defined a new information age in our society. Microblogging sites, such as Twitter, gained immense popularity during the COVID-19 pandemic. We developed an automated framework to extract the positive, negative, and neutral sentiments from tweets, and to further classify the tweets through machine-learning (ML) techniques. The developed framework can help to understand the sentiments in our society during profound events, such as the COVID-19 pandemic. Our framework is novel in that it is a hybrid framework that combines a lexicon-based technique for tweet sentiment analysis and labeling with supervised ML techniques for tweet classification. We have evaluated the hybrid framework with a range of measures, such as precision, accuracy, recall, and F1 score. Our results indicate that the majority of the sentiments are positive (38.5%) or neutral (34.7%). Furthermore, with an accuracy of 83%, the long short-term memory (LSTM) neural network has been selected as the preferred ML technique in the framework. The evaluation results indicate that our hybrid framework has the potential to automatically classify large tweet volumes, such as the tweets on COVID-19, according to the sentiments in the society.