This repository contains source code, data, and results for the paper "Simulating Diverse Patient Populations Using Patient Vignettes and Large Language Models", presented at the CL4Health Workshop at LREC-COLING 2024.
./data/input
: Contains the pre-defined patient vignettes../data/output
: Contains the results of the model predictions, comparing the answers based on each of the three vignettes.
- Install pdm on your machine.
- Run
pdm install
to install the dependencies. - Create an account at OpenAI and get an API key.
- Run
pdm run-app
to spin-up a local server.
- Open your browser and navigate to
http://localhost:8501
. - Enter the API key you received from OpenAI.
- Select one of the two OpenAI models. More details to be found here.
- Select one of the pre-defined patient vignettes or create your own.
- Interact with the model.
- Export dependencies with
pdm export -o requirements.txt --without-hashes
(Streamlit does currently not support pdm). - Build docker image.
- Run docker container. Server port: 8501.