Text Clustering of COVID-19 Vaccine Tweets

Abstract

The advent of the novel coronavirus disease (COVID-19) in late December 2019 led to the dramatic loss of human life worldwide and presented an unprecedented challenge to public health, education, social life, world economics, and the world of work. Equal access to safe and effective vaccines is very vital to ending the coronavirus pandemic. This research paper aims to perform text clustering on COVID-19 vaccine tweets. It investigates the optimal number of clusters prevalent in the COVID-19 vaccine corpus using deep learning techniques and machine learning algorithms. The study also investigates how using word embeddings can improve the accuracy of the proposed models by evaluating unsupervised learning methods. Machine learning clustering algorithms such as k-means and HDBSCAN, deep learning-based clustering techniques, and UMAP a dimensionality reduction algorithm were employed to perform text clustering. The results of this research showed the optimal clusters obtained by using deep learning clustering techniques and machine-learning algorithms for text clustering. HDBSCAN clustering algorithm showed better clustering results based on features learned while k-means performed better clustering based on various evaluation metrics.