https://github.com/gforge/medbench-dataprep

Preparation of data for the MedBench project
https://github.com/gforge/medbench-dataprep

Last synced: 17 days ago
JSON representation

Preparation of data for the MedBench project

• Host: GitHub
• URL: https://github.com/gforge/medbench-dataprep
• Owner: gforge
• License: mit
• Created: 2024-08-01T13:56:06.000Z (3 months ago)
• Default Branch: main
• Last Pushed: 2024-09-30T20:41:45.000Z (about 1 month ago)
• Last Synced: 2024-10-03T15:13:59.556Z (about 1 month ago)
• Language: R
• Size: 64.5 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# MedBench-DataPrep

## Overview

Welcome to the MedBench-DataPrep repository. This repository contains R functions necessary for generating data for the MedBench research study. The study aims to create a benchmark dataset of fictional Electronic Health Records (EHRs) to evaluate and improve the performance of Large Language Models (LLMs) in medical documentation tasks.

## Study Objectives

- **Dataset Creation**: Develop a comprehensive dataset of fictional EHRs, covering a wide range of medical scenarios.

- **Benchmarking**: Establish benchmarks to assess LLM performance in handling medical jargon, demographic diversity, and factual inconsistencies.

- **Evaluation**: Implement methods to quantitatively and qualitatively evaluate the generated medical documentation.

## Methods

- **EHR Structure**: Each case includes core medical notes and related lab and medication data.

- **Dataset Composition**: The dataset will reflect standard medical scenarios and include variations to test LLM capabilities under different conditions.

- **Prompt Methodology**: Explore advanced prompting techniques to enhance LLM performance in generating medical documentation.

## Contributing

We welcome contributions from the community. Please refer to our contribution guidelines for more information.

## License

This project is licensed under the MIT License. See the LICENSE file for details.