Big Data-Driven Social Media Analytics Using a COVID-19 Dataset

Social media continues to proliferate rapidly and generate substantial volumes of usergenerated content. In the field of COVID-19, user-generated data on Twitter can be utilized by policymakers, researchers, and healthcare professionals for seeking to understand public sentiment and behavior in real time. This is imperative to guarantee more informed, timely, and targeted interventions during a public health crisis and to prepare for future pandemics in an increasingly skeptical world with a swiftly growing population. It gives unprecedented opportunities to recognize the behavioral patterns and opinions of social media users with regards to COVID-19. Our study presents a big data-driven approach for analyzing a large dataset of tweets associated with COVID-19. We utilize modern natural language processing techniques and approaches, including but not limited to transformer-based embeddings, clustering, and zero-shot classification, we automatically extract latent topics, sentiments, bot-generated tweets and outliers. The framework further provides actionable insights from the dataset. We aim to demonstrate that large-scale analysis of COVID-19 tweets has the potential to reveal distinctive patterns which are useful for research and formulation of public policy. Our goal is to highlight the potential of big data analytics to transform semi-structured textual data into insightful and actionable information.