Sentiment Analysis Methods for HPV VaccinesRelated Tweets Based on Transfer Learning

Area of Science:

• Computational Social Science
• Public Health Informatics
• Machine Learning Applications

Background:

• Social media generates vast data for public sentiment analysis, vital for understanding human papillomavirus (HPV) vaccine coverage.
• Unannotated social media data and the high cost of annotation limit deep learning applications for sentiment analysis.
• Understanding public opinion on HPV vaccines is essential for addressing low vaccination rates.

Purpose of the Study:

• To propose and evaluate transfer learning approaches for analyzing public sentiment towards HPV vaccines on Twitter.
• To overcome the challenge of limited annotated data in social media sentiment analysis.
• To identify effective strategies for improving HPV vaccine uptake through sentiment analysis.

Main Methods:

• Three transfer learning methods were applied: DWE-BiGRU-Att (transferring static and ELMo embeddings), fine-tuning generative pre-training (GPT), and fine-tuning BERT.
• The models were trained and evaluated on a Twitter dataset related to HPV vaccination.
• Comparative analysis of the performance of different transfer learning techniques.

Main Results:

• All three proposed transfer learning methods demonstrated efficacy in sentiment analysis for the HPV vaccination task.
• The fine-tuned BERT model achieved superior performance compared to other evaluated methods.
• The study successfully applied machine learning to analyze public sentiment on HPV vaccines from social media.

Conclusions:

• Transfer learning, especially fine-tuned BERT, is a powerful approach for analyzing public sentiment on HPV vaccines using social media data.
• These findings can guide the development of targeted interventions to enhance vaccine acceptance and coverage.
• Effective sentiment analysis of social media data can provide valuable insights for public health initiatives.