Aspect Based Twitter Sentiment Analysis on Vaccination and Vaccine Types in COVID-19 Pandemic With Deep Learning.

Area of Science:

• Public Health
• Computational Social Science
• Vaccinology

Background:

• COVID-19 vaccine development and rollout faced societal vaccine opposition and mistrust, hindering vaccination campaigns.
• Understanding public sentiment towards vaccines is crucial for effective public health strategies during pandemics.

Purpose of the Study:

• To conduct aspect-based sentiment analysis of public opinion on COVID-19 vaccination and vaccine types using Twitter data.
• To investigate and compare public sentiment across different countries (USA, UK, Canada, Turkey, France, Germany, Spain, Italy).

Main Methods:

• Collected and processed 928,402 vaccine-focused tweets in English and Turkish.
• Employed aspect-based sentiment analysis using four BERT models (mBERT-base, BioBERT, ClinicalBERT, BERTurk) across four aspects: policy, health, media, and other.
• Performed sentiment analysis on tweets related to the six most frequent COVID-19 vaccines.

Main Results:

• Achieved high performance with an F1 Score between 84% and 88% for country-specific datasets and an overall accuracy of 87%.
• Presented country-specific insights into public views on vaccination and vaccine types.
• Identified key aspects influencing public sentiment towards COVID-19 vaccines.

Conclusions:

• The study provides valuable insights into public perception of COVID-19 vaccines, crucial for addressing vaccine hesitancy.
• The developed methodology demonstrates effectiveness in analyzing large-scale social media data for public health insights.
• Findings can inform targeted communication strategies to improve vaccine acceptance and uptake.