Opinion classification at subtopic level from COVID vaccination-related tweets

Area of Science:

• Public Health
• Computational Social Science
• Infectious Disease Epidemiology

Background:

• Coronavirus disease 2019 (COVID-19) is a global pandemic with high mortality.
• Vaccination is a key strategy to control the spread of COVID-19.
• Public opinion on COVID-19 vaccination varies widely and is often expressed on social media.

Purpose of the Study:

• To classify public opinions on COVID-19 vaccination at a subtopic level.
• To analyze sentiments (positive, negative, neutral) expressed in tweets regarding COVID-19 vaccination.
• To develop a robust model for real-time sentiment analysis of public discourse on vaccination.

Main Methods:

• Utilized opinion mining and sentiment analysis techniques.
• Employed Rocchio query expansion to generate query sets for sentiment classification.
• Applied Latent Dirichlet Allocation (LDA) algorithm to identify twenty distinct subtopics within tweets.
• Integrated Apache Kafka for real-time tweet classification.

Main Results:

• Developed an LDA model with a coherence score of 0.56, identifying twenty relevant subtopics.
• Successfully classified tweets into positive, negative, and neutral sentiment categories based on identified subtopics.
• Demonstrated the capability for real-time classification of public opinion on COVID-19 vaccination.

Conclusions:

• The proposed LDA-based model effectively classifies public opinions on COVID-19 vaccination into nuanced subtopics.
• Real-time analysis of public sentiment regarding vaccination is feasible and valuable for public health strategies.
• Understanding subtopic-level sentiment provides deeper insights into public response to vaccination campaigns.