ABSTRACT
In this paper, we provide a sentiment analysis of conversations surrounding Covid-19 vaccine adoption on Twitter. We focus on key regions of the US, particularly urban areas with high African American populations. We utilize machine learning models such as logistic regression, Support Vector Machines, and Naive Bayes to provide baseline models. Furthermore, we develop fined-tuned Transformer-based language models that provide a classification of sentiments with high accuracy. The results from our analysis show that fine-tuning our dataset on a Transformer-based model, Covid-BERT v2, performs better than our baseline models however the accuracy is still relatively low. This might be as a result of the very limited training dataset. Future work will explore the use of a higher quality dataset and also evaluate other transformer-based models. © 2021 IEEE.