Utilizing Large Language Models to Monitor Social Media for Disability: An Analysis of Sentiment and Disability Models in Tweets.

Journal: Studies in health technology and informatics
Published Date:

Abstract

This study explores how well large language models (like the kind that powers ChatGPT) can analyze online conversations about disability rights. We specifically looked at whether these models could: 1) identify if tweets about people with disabilities were positive or negative, and 2) tell if the tweets viewed disability as a problem with society (social model) or a problem with the individual (medical model). We collected 5,000 tweets and trained a language model to analyze them. The results demonstrated promising accuracy levels for sentiment analysis and social vs. medical model classification.

Authors

  • Abdul Hamid Dabboussi
    Lassonde School of Engineering, York University, Toronto, Canada.
  • Iman Yousuf
    School of Information Technology, York University, Toronto, Canada.
  • Hannah Bullock
    School of Health Policy and Management, York University, Toronto, Canada.
  • Naleni Jacob
    School of Health Policy and Management, York University, Toronto, Canada.
  • Vijay Mago
    Department of Computer Science, Lakehead University, 955 Oliver Road, Thunder Bay, P7B 5E1, Canada. vmago@lakeheadu.ca.
  • Christo El Morr
    York University, School of Health Policy and Management, Canada.