LLMs as Sentiment Analysis Experts
Evaluating AI accuracy in tobacco product sentiment analysis
This study evaluates GPT-3.5 and GPT-4 Turbo's ability to match human expert sentiment analysis of heated tobacco products on social media.
- GPT models assessed 1,000 social media messages about tobacco products
- Compared AI classifications (pro, anti, neutral) with human expert evaluations
- Demonstrated LLMs' potential to streamline labor-intensive sentiment analysis for public health research
Why it matters: This research bridges medical and data science fields by providing efficient tools for monitoring public sentiment about tobacco products, enabling faster public health interventions and policy development.