
Healthcare ARS System
- Supervisor:Hee Choel Kim
- Status: Upcoming
Common approach -> Training Feature-based machine learning models or fine-tuning large language models. But Existing health-related question-answering datasets are either * Unlabeled, * Relatively small * Do not cover health-related questions. Yes/No Questions * Yes/No Questions are more prone to misinformation * Illustrate the vulnerability of such questions to misinformation and unfairness.
Lack of Context.
Collect a large data set of questions, automatically labelling them into three categories: medical, health-related, and non-health-related.
Propose a new method to identify health-related questions
Build a new Question/Answering system using specific literature to infer the correct answer to a yes/no question and
Also, find an authentic web document supporting this answer correctly.