Poster

JSON Data Generation: Linking Statistical Analysis with Large Language Models

Changhong Shi, Jaime Yan

PHUSE/FDA Computational Science Symposium (CSS) 2024 · 2024 · PP02

Abstract

This poster presents a framework that extracts schema from structured clinical datasets (e.g., CDISC ADaM) into JSON, uses it to design a graph-based database (Neo4j) with graph embeddings stored in a vector database, and enables LLMs to answer natural-language questions by retrieving relevant variables and generating SQL or Cypher queries — automating clinical data analysis workflows.

Keywords

JSON · CDISC ADaM · Neo4j · Cypher · LLM · vector database

Cite

Changhong Shi, Jaime Yan. "JSON Data Generation: Linking Statistical Analysis with Large Language Models." PHUSE/FDA Computational Science Symposium (CSS) 2024, 2024. doi:10.5281/zenodo.22182906.