This repository contains a conceptual model and generated dataset as presented in the publication "Synthetic Ground Truth for Evaluating Generative Policy Models".
A Docker image is provided for quick setup.
docker build -t coalition-model .
docker run -it -p 8080:8080 coalition-model:latest
This runs a Controlled English store accessible via the URL: http://localhost:8080/ce-store/ui. Navigate to this URL and load the DAIS-ITA local sentence set.
Data can then be generated by running the generate_data.py script within the scripts folder. This script accepts the following command line options:
--granularity G number of values between min and max to generate for each
environmental condition. Default 2
--num_assets A number of assets to generate per coalition partner per
mission. Default 2
--bb_string B swlon,swlat,nelon,nelat bounding box for asset generation.
--decision D boolean condition used to evaluate approve/reject
--num_requests R number of requests to generate. Default 100