Sentiment Analysis for COVID Vaccinations Using Twitter: Text Clustering of Positive and Negative Sentiments

The objective of the chapter is to analyze Twitter data to extract sentiments and opinions in unstructured text. The team attempted to use contextual text analytics to categorize Twitter data to understand the positive or negative sentiments for COVID vaccinations and wish to highlight key concerns. Text clustering has been performed on positive and negative sentiments to understand the key themes behind them. We followed a two-step process. In the first step, we identified positive and negative sentiments from Twitter feeds. In the second step, we aggregated all sentiments into categories to deduce what the Twitterati is thinking about COVID-19 vaccinations. The whole analysis was performed using Python, including TextBlob and Vader libraries. TextBlob library uses the Naïve-Bayes (probabilistic algorithms using Bayes’s Theorem to predict the category of a text) classifier to assess the polarity of a sentence and generates a score ranging between −1 (strongly negative) and +1 (strongly positive). The Naïve Bayes classifier categorizes based on probabilities of events. Although it is a simple algorithm, it performs well in many text classification problems. On the other hand, the Vader library uses a lexical approach that uses preassigned scores labeled positive and negative for different words found in a text. These scores are based on pre-trained models classified as positive/negative by actual human beings. We then performed the topic extraction that discovers the keywords in sentiments that capture the recurring theme of a text and is widely used to analyze large sets of sentiments to identify the most common topics easily and efficiently. We found a large segment as neutral (53%) followed by a positive sentiment segment that contributed 36% of tweets. However, at the same time, many people (10%+) remain on the fence regarding the potential repercussions of COVID vaccines as they are relatively new and yet untested over longer periods of time. It is reasonable to expect that people are a bit skeptical about vaccinations. Text clustering of negative sentiments identified late vaccinations and side effects being the key concerns. Positive sentiments mainly were driven by the readiness of other vaccines and weak reactions following vaccinations. The study contributes to text mining literature by providing a framework for analyzing public sentiments. This can help to understand the key themes in negative sentiments related to COVID vaccinations and can help in adjusting policies.