Correction: COVID-19 Vaccine Tweets After Vaccine Rollout: Sentiment–Based Topic Modeling

Luwen Huangfu, Yiwen Mo, Peijie Zhang, Daniel Zeng, Saike HeJournal of Medical Internet Research 2022 · DOI · OpenAlex

Abstract

This article provides a formal correction to the study "COVID-19 Vaccine Tweets After Vaccine Rollout: Sentiment–Based Topic Modeling" (DOI: 10.2196/31726). The correction addresses a critical visualization error in the sentiment tracking for the Moderna vaccine.

Introduction

Accurately tracking public discourse during a global pandemic is essential for public health officials to combat vaccine hesitancy and misinformation. This study utilized sentiment-based topic modeling to analyze Twitter data following the initial rollout of major COVID-19 vaccines. However, a post-publication review identified a color-coding error in Figure 8, specifically affecting the Moderna vaccine data. This correction ensures the integrity of the findings regarding public perception.

Methodology and Data

The research analyzed a longitudinal dataset of tweets from December 14, 2020, to April 30, 2021. The methodology involved sentiment analysis to calculate daily average positive and negative sentiment scores, topic modeling to identify themes within the discourse, and trend smoothing to highlight long-term shifts. The study categorized data by manufacturer: Johnson & Johnson, AstraZeneca, Pfizer, and Moderna.

Corrected Findings: Moderna Sentiment

In the original publication, the positive and negative sentiment lines for the Moderna vaccine were swapped in Figure 8D. The corrected visualization confirms that positive sentiment (blue line) consistently remained higher than negative sentiment (red line).

Corrected Moderna Daily Average Sentiment Score showing the dominance of positive sentiment.

The corrected data shows a stable positive sentiment trend for Moderna, hovering around a score of 0.10, while negative sentiment remained lower and relatively flat, near 0.05. This aligns with the trends observed for other mRNA vaccines like Pfizer, suggesting a robust public confidence in the technology during the early rollout.

While the overall discourse remained net-positive across all brands, the raw daily scores revealed significant volatility. These spikes often correlate with real-world news events, such as clinical trial updates or regulatory reports, demonstrating the sensitivity of public opinion to the 24-hour news cycle.

AstraZeneca Daily Average Sentiment Score illustrating the volatility of public sentiment in response to news events.

For instance, AstraZeneca showed a more pronounced dip in positive sentiment and a slight rise in negative sentiment toward the end of April 2021. This likely reflects the emergence of global news regarding blood-clotting concerns that occurred during that specific timeframe.

Conclusion

This correction underscores the importance of visualization integrity in data science. By rectifying the color-coding error, the authors confirm that public sentiment toward the Moderna vaccine was overwhelmingly positive during the initial rollout period, mirroring the broader trend of positive vaccine discourse on social media during early 2021.

Summary updated

Abstract

[This corrects the article DOI: 10.2196/31726.].
BibTeX
@article{huangfu2022correction,
  author = {Luwen Huangfu and Yiwen Mo and Peijie Zhang and Daniel Zeng and Saike He and Daniel Dajun Zeng},
  title = {{Correction: COVID-19 Vaccine Tweets After Vaccine Rollout: Sentiment–Based Topic Modeling}},
  journal = {Journal of Medical Internet Research},
  year = {2022},
  doi = {10.2196/37841},
}
Figure