Transformer-based LLMs
Date:
Transforming Social Media data into Mental health Insights: Decoding Drug Abuse using LLMs
- Applied Machine Learning Research in Public Health Domain to build a framework for studying substance usage using social media data on a very large scale.
- Developed a pipeline for preprocessing large scale social media data; data cleaning, data mining other features
- Employed a RoBERTa model by pre-training and fine tuning.
- Challenged by limited annotated data, the research involved iterative learning mechanism with Human-in-The-Loop (HiTL)
- Detailed qualitative and quantitative analysis of substance Use trend and patterns focusing in COVID-19 pandemic period.
- Aligned the research findings with the national survey based reports like NCDAS and MTA, providing a baseline substance use reports from social media perspective.