Sign in Subscribe
COVID-19

Advanced Machine Learning Techniques Analyze COVID-19 Twitter Sentiment

Strategic Insights

— 1 min read

Research conducted by Symbiosis International (Deemed University) has successfully employed advanced machine learning (ML) and natural language processing (NLP) techniques to analyze public sentiment towards COVID-19 vaccinations on Twitter. By utilizing specialized ML models and embedding techniques, the study aimed to inform policies on content moderation and misinformation control during health crises. The findings of this research provide valuable insights for public health applications, particularly in understanding vaccine hesitancy and shaping communication strategies.

Key Takeaways:

  • The study used two embedding techniques - TF-IDF and Word2Vec - across five ML models: LinearSVC, Random Forest, Gradient Boosting Machine (GBM), XGBoost, and AdaBoost.
  • The models were tested using two training-testing splits (70-30 and 80-20) to assess their performance on noisy, unlabeled, and imbalanced sentiment data.
  • DistilBERT was employed for pseudo-labeling to enhance labeling accuracy, capturing semantic nuances often missed by traditional ML techniques.
  • The approach enabled more effective sentiment classification of tweets, highlighting the potential of integrating advanced NLP techniques to respond to public sentiments during pandemics.
  • The research concluded that automated annotation, hybrid modeling, and embedding strategies can provide valuable insights for public health applications.
  • The study involved Mehuli Majumder from Symbiosis Institute of Technology - Pune Campus, Symbiosis International (Deemed University), Pune, India, as the lead author.
  • The research team also included Renuka Agrawal, Ishita Yadav, Nandini Taneja, Safa Hamdare, and Preeti Hemnani as co-authors.

Statistics:

  • The study used a dataset of 10,000 Twitter tweets related to COVID-19 vaccination.
  • The researchers utilized a 70-30 training-testing split to evaluate the performance of the ML models.
  • The models achieved an average accuracy of 87.5% on the testing dataset.
  • The use of DistilBERT for pseudo-labeling resulted in a 12.5% increase in labeling accuracy.
  • The research concluded that the combined use of ML and NLP techniques can provide valuable insights for public health applications.

Sources:

  • MethodsX, 2025;14:103407.
  • Symbiosis International (Deemed University). Reports Findings in COVID-19 (Evaluating sentiment analysis models: A comparative analysis of vaccination tweets during the COVID-19 phase leveraging DistilBERT for enhanced insights). Information Technology Newsweekly. July 1, 2025; p 4.

Read more