AUTHOR=Jalil Zunera , Abbasi Ahmed , Javed Abdul Rehman , Badruddin Khan Muhammad , Abul Hasanat Mozaherul Hoque , Malik Khalid Mahmood , Saudagar Abdul Khader Jilani 

TITLE=COVID-19 Related Sentiment Analysis Using State-of-the-Art Machine Learning and Deep Learning Techniques

JOURNAL=Frontiers in Public Health

VOLUME=Volume 9 - 2021

YEAR=2022

URL=https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2021.812735

DOI=10.3389/fpubh.2021.812735

ISSN=2296-2565

ABSTRACT=The Coronavirus (COVID-19) pandemic has influenced the everyday life of people around the globe. In general and during lockdown phases, people worldwide use social network media to state their viewpoints and general feelings concerning the pandemic that has hampered their daily lives. Twitter is one of the most commonly used social media platforms. It showed a massive increase in tweets related to Coronavirus, including positive, negative, and neutral tweets, in a minimal period. The researchers move towards the sentiment analysis and analyze the various emotions of the public towards COVID-19 due to the diverse nature of tweets. The conventional sentiment analysis methods only identify the polarity and classify it in positive, negative, or neutral tweets. As an advanced step, in this paper, our proposed framework analyzes COVID-19 by focusing on people who share their opinions on social media networking sites such as Twitter. The proposed framework analyzes collected tweets' sentiments for sentiment classification using various features sets and classifiers. The early detection of COVID-19 sentiments from collected tweets allows the better handling of the current pandemic situation. Tweets are classified into positive, negative, and neutral sentiment classes. Experiments were conducted to evaluate the performance of Machine Learning (ML) and Deep Learning (DL) classifiers using different evaluation metrics. According to the experimental analysis, the proposed approach provides consistent performance for the COVIDSenti dataset. Experiments prove that the proposed Multi-depth DistilBERT provides better accuracy of 96.66%, 95.22%, 94.33%, and 93.88% for COVISenti, COVIDSenti_A, COVIDSenti_B, and COVIDSenti_C, respectively, which is better accuracy among all other methods used in this study as well as compared to previous state-of-art approaches and traditional machine learning and deep learning algorithms.