1. Open source Extraction Exploration

Once we have LLM-generated medical notes we then want to extract entities from these notes to then produce a privacy risk score.

In previous work, AWS Comprehend Medical was first used to extract entities from these medical notes. In this project we want to explore using open-source named-entity extraction methods that could be used instead of AWS Comprehend Medical.

Experiments and example notebooks for the extraction component of PrivFp are available in the privfp-experiments repository.