PubMed Case-Report Corpus for Clinical Narrative Privacy Research

Description

This repository contains the datasets used to investigate reconstruction risk from fragmented clinical narratives and to evaluate structure-aware anonymization strategies. The collection consists of PubMed case-report abstracts obtained through predefined literature searches and exported in CSV format.

The repository includes three datasets used for development of the reconstruction framework (interstitial cystitis/bladder pain syndrome, vulvodynia, and fibromyalgia) together with independent datasets used for external validation (endometriosis and irritable bowel syndrome). Each CSV file contains bibliographic information and abstract text extracted from PubMed records. No additional annotation, preprocessing, or manually curated labels are included in this repository.

The datasets were assembled to support reproducible analyses of clinical fingerprint extraction, fragment-based patient reconstruction, sensitive-information leakage assessment, and anonymization benchmarking. They may also be useful for research on clinical natural language processing, information retrieval, privacy-preserving data sharing, biomedical text mining, and evaluation of de-identification methods.

The data originate from publicly available PubMed records. Users are responsible for complying with the applicable terms of use and citation requirements of the original data sources. This repository distributes only the compiled research dataset and does not claim ownership of the underlying PubMed records.