Yale NLP/LLM Interest Group Talks

Translating complex eligibility criteria into SQL queries is a crucial task for clinical trials due to its efficiency in finding eligible patients. This process often requires significant human effort, making it both time-consuming and prone to errors. I will first review the evolution of parsing clinical trial criteria, focusing on methods before and after the advent of Large Language Models (LLMs) and how LLMs have transformed this process. Then, I will highlight current challenges, including issues in LLM-based SQL generation, managing intricate logical relationships, and interpreting complex phenotypes. Following this, I will introduce our current work aimed at addressing these challenges through a novel pipeline designed to leverage LLMs in parsing criteria and improve their performance in generating SQL queries.