Modeling Vaccine Discourse and Polarization on YouTube in Brazil: A Semi-Supervised Stance Detection Approach

• Geovana S. de Oliveira UFOP
• Carlos H. G. Ferreira UFOP

DOI: https://doi.org/10.5753/sbcas_estendido.2026.21609

Resumo

Vaccination is a cornerstone of public health, yet the COVID-19 pandemic exposed how online misinformation and polarization can undermine immunization efforts. Studying these dynamics at scale is challenging due to scarce annotated data, class imbalance, and high labeling costs. This work presents a longitudinal analysis of vaccine discourse on YouTube in Brazil (2018–2024), using a semi-supervised stance detection framework to classify 1.4 million comments. The approach enables effective learning from limited supervision and produces a publicly available Portuguese stance classification model alongside a large-scale analysis of online debate. Results show strong performance despite class imbalance and enable the analysis of polarization over time. We find that polarization intensified during the pandemic and became more fragmented afterward, with science communication and digital-native channels concentrating both supportive and opposing engagement.

Referências

Abao, R. P., Estuar, M. R. J. E., Cataluña, A. A. M., Aureus, J. P., and Mapua, D. C. (2021). Emotion analysis of comments from vaccine-related youtube videos: Understanding the public’s response to covid-19 vaccination. In IEEE SNAMS, pages 1–7.

Amini, M.-R., Feofanov, V., Pauletto, L., Hadjadj, L., Devijver, E., and Maximov, Y. (2025). Self-training: A survey. Neurocomputing, 616:128904.

Aygün, I., Kaya, B., and Kaya, M. (2021). Aspect based twitter sentiment analysis on vaccination and vaccine types in covid-19 pandemic with deep learning. IEEE J-BHI, 26(5):2360–2369.

Blanco, G., Yáñez Martínez, R., and Lourenço, A. (2025). Leveraging deep learning to detect stance in spanish tweets on covid-19 vaccination. JAMIA open, 8(1):ooaf007.

Dubey, A., Jauhri, A., Pandey, A., Kadian, A., Al-Dahle, A., Letman, A., Mathur, A., Schelten, A., Yang, A., Fan, A., et al. (2024). The llama 3 herd of models. arXiv e-prints, pages arXiv–2407.

Fleiss, J. L., Levin, B., and Paik, M. C. (1981). The comparison of proportions from several independent samples. Statistical methods for rates and proportions.

Knisely, J. M. and Erbelding, E. (2025). Vaccines for global health: Progress and challenges. The Journal of Infectious Diseases.

Larson, H. J. (2022). Defining and measuring vaccine hesitancy. Nat. Hum. Behav., 6(12):1609–1610.

Locatelli, M. S., Caetano, J., Meira Jr, W., and Almeida, V. (2022). Characterizing vaccination movements on youtube in the united states and brazil. In ACM HT.

Miyazaki, K., Uchiba, T., Kwak, H., An, J., and Sasahara, K. (2024). The impact of toxic trolling comments on anti-vaccine youtube videos. Sci. Rep., 14(1):5088.

Pan, Y., Wang, Q., Yang, P., Zhang, L., Wu, S., Zhang, Y., Sun, Y., Duan, W., Ma, C., Zhang, M., et al. (2017). Influenza vaccination in preventing outbreaks in schools: A long-term ecological overview. Vaccine.

Pércio, J., Fernandes, E. G., Maciel, E. L., and Lima, N. V. T. d. (2023). 50 years of the brazilian national immunization program and the immunization agenda 2030. Epidemiologia e Serviços de Saúde, 32:e20231009.

Qorib, M., Oladunni, T., Denis, M., Ososanya, E., and Cotae, P. (2023). Covid-19 vaccine hesitancy: Text mining, sentiment analysis and machine learning on covid-19 vaccination twitter dataset. Expert Syst. Appl.

Song, M. Y.-J. and Gruzd, A. (2017). Examining sentiments and popularity of pro-and anti-vaccination videos on youtube. In Social Media + Society, pages 1–8.

Straton, N. (2023). Covid vaccine stigma: detecting stigma across social media platforms with computational model based on deep learning. Appl. Intell., 53(13):16398–16423.

Sufi, F. K., Razzak, I., and Khalil, I. (2022). Tracking anti-vax social movement using ai-based social media monitoring. IEEE-TTS.

Zhao, S., Hu, S., Zhou, X., Song, S., Wang, Q., Zheng, H., Zhang, Y., Hou, Z., et al. (2023). The prevalence, features, influencing factors, and solutions for covid-19 vaccine misinformation: systematic review. JPHS, 9(1):e40201.