Work

MedDeID English synthetic clinical corpus, benchmark and annotation guideline

Year: 2026

Type: dataset

Abstract: A synthetic English clinical-text de-identification release containing 6,700 model-development documents and a disjoint 300-document human-validated benchmark, balanced between en-GB and en-US. The benchmark contains 1,717 primary PII spans and 7,358 confirmed nested core-PII subannotation segments. The archive also contains the English annotation guideline and provenance manifests. It contains no real patient notes or personal information.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.22127863

Authors: Hellemans, Stig, Stroobants, Tom, Scheurwegs, Elyne, Meysman, Pieter, Jorens, Philippe, Laukens, Kris

Institutions: University of Antwerp, Antwerp University Hospital

Language: English

Topic: Privacy-Preserving Technologies in Data

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

Funder: Fonds Wetenschappelijk Onderzoek

Award: QUEHR.AI: Towards a Question-Answering System for Electronic Health Records: Bridging Clinicians and Data

OpenAlex ID: W7204493575

Cites:

Cited by: 1

Related to: 10