JMIR Publications

ABSTRACT

Background:

COVID-19 vaccines are considered as one of the most effective preventive strategies for containing the pandemic. Having a better understanding of the public’s conceptions of COVID-19 vaccines could help the effort to vaccinate the community promptly and thoroughly. However, no known empirical research has fully explored the public’s vaccine awareness by a sentiment-based topic modeling approach and therefore little is known about the evolution of public attitude since the rollout of COVID-19 vaccines.

Objective:

In this paper, we focus specifically on the tweets about COVID-19 vaccines after they became publicly available. We aim to explore the overall sentiments and topics about COVID-19 vaccines, as well as how the sentiments and main concerns evolve.

Methods:

We collected tweets related to COVID-19 vaccines from December 8th, 2020, to April 8th, 2021, using Twitter API, resulting in over 158,000 tweets after data cleaning. We applied sentiment-based topic modeling to our dataset. For the topic modeling, we used the coherence score to determine the optimal topic number and calculated the topic distribution to illustrate the topic evolution.

Results:

There were overall 38.90% positive, 17.65% negative, 41.26% neutral, 1.80% highly positive, and 0.39% highly negative sentiments among these tweets. Our results revealed that the main topics of positive tweets were (a1) happiness with the expectation of getting the vaccine (38.65%), (a2) recognition of the importance to get vaccinated (11.38%), and (a3) thankfulness to the healthcare staff’s efforts (10.95%); while the main concerns underlying negative tweets were (b1) the side effects of vaccines (36.85%), (b2) the government’s power abuse (12.00%), and (b3) fear of adverse events, including death (9.82%). The negative sentiment was volatile and could be easily influenced; however, it became gradually weaker, presumably because of the encouraging signs of the effectiveness of COVID-19 vaccines. We also demonstrated how the main concerns had changed (via topic heatmap visualization) during the widespread vaccination campaign.

Conclusions:

To the best of our knowledge, it is the first study to evaluate the public’s COVID-19 vaccine awareness via sentiment-based topic modeling on social media since its rollout. This study builds upon a text mining framework combining sentiment analysis and topic modeling that automatically captures and analyzes COVID-19 vaccine-related discussions on social media, enabling real-time assessments of the public’s vaccine awareness. Our result could help policymakers and research communities to track public attitudes on COVID-19 vaccines and aid their decision-making in promoting the vaccination campaign.