Vaccine-Hesitancy-NLP

Public acceptance toward the COVID-19 vaccines are critical factors for governments looking to curtail the spread of the virus and accelerate economic recovery through vaccination programs. Vaccine hesitancy is increasing due to the proliferation of fake news sources and a general anti-science sentiment, and due to justifiable concerns over the vaccines’ rapid developmental and testing phase. To inform successful vaccination campaigns, we provide governments with a visual tool to help track vaccine hesitancy over time and geography. This visual tool condenses the public’s opinion using natural language processing (NLP) techniques applied to a comprehensive dataset of Tweets related to COVID-19 vaccination. Topic modelling and sentiment analysis techniques are applied to a data pipeline to produce reliable information regarding users’ general sentiment towards vaccines and the themes and topics associated with those sentiments. We developed a data pipeline, developed filtering, preprocessing and text vectorization techniques, as well as topic modelling and sentiment analysis models leveraging BERT. Additionally, we provided performance benchmarks to BERT architectures by using Multinomial Naive Bayes, Logistic Regression and Latent Dirichlet Allocation. Data visualizations using Plotly were also created to display our findings.