careons/dutch-healthcare-pii-ner

Dutch Healthcare PII NER Dataset

A synthetic Dutch-language Named Entity Recognition (NER) dataset focused on Personally Identifiable Information (PII) detection and anonymization, with a strong emphasis on healthcare contexts in the Netherlands.

This dataset is fully synthetic. All texts and entities were generated by AI and do not represent real individuals, organizations, or medical records.

---

Overview

Property	Value
Language	Dutch (nl)
Samples	400
Total annotated entities	2,065
Avg. entities per sample	5
Entity types	19
Domain	Healthcare, insurance, municipal records, general correspondence
Generation method	Synthetic (LLM-generated)
Task	Token classification / Named Entity Recognition

---

Motivation

This dataset was created to support the development of a Dutch PII detection and anonymization system as part of a health informatics project. Dutch healthcare records, insurance documents, and patient correspondence contain a wide variety of sensitive personal identifiers — from standard PII (names, addresses) to Dutch-specific identifiers such as BSN numbers and health insurance policy numbers (ZORGPOLIS_NUMBER).

No suitable large-scale annotated Dutch PII dataset with healthcare coverage was publicly available, so this synthetic dataset was generated to fill that gap.

---

Entity Types

The dataset covers 19 entity types, including Dutch-specific and healthcare-specific identifiers:

Entity Type	Description	Example
PERSON	Full name of an individual	Jan de Vries
LOCATION	City, address, building name, or geographic reference	Amsterdam, LUMC
DATE	Calendar date	14 maart 2023
TIME	Time of day	08:30
PHONE_NUMBER	Dutch phone number	+31 6 12345678
EMAIL_ADDRESS	Email address	j.devries@email.nl
URL	Web address	https://www.mijnzorgportaal.nl
ZIPCODE	Dutch postal code	1234 AB
GPS_COORDINATES	Geographic coordinates	52.3676° N, 4.9041° E
IBAN_CODE	Dutch/European IBAN bank account number	NL91ABNA0417164300
CREDIT_CARD	Credit card number (Luhn-valid)	4111 1111 1111 1111
CVV	Card verification value	372
BSN	Dutch Citizen Service Number (Burgerservicenummer) — passes 11-proof	123456782
PASSPORT	Passport number	NX1234567
ZORGPOLIS_NUMBER	Dutch health insurance policy number	12345678
IP_ADDRESS	IPv4 or IPv6 address	192.168.1.1
MAC_ADDRESS	Network hardware address	00:1A:2B:3C:4D:5E
IMEI	Mobile device identifier (Luhn-valid)	490154203237518
LICENCE_PLATE	Dutch vehicle licence plate	XK-98-LM

Entity Distribution

PERSON           391  ██████████████████████
LOCATION         247  ██████████████
ZIPCODE          136  ████████
TIME             134  ████████
BSN              128  ████████
DATE             127  ███████
GPS_COORDINATES  102  ██████
CREDIT_CARD       94  █████
PHONE_NUMBER      92  █████
EMAIL_ADDRESS     82  █████
ZORGPOLIS_NUMBER  79  ████
IBAN_CODE         77  ████
LICENCE_PLATE     75  ████
URL               75  ████
CVV               71  ████
MAC_ADDRESS       45  ███
IP_ADDRESS        41  ██
PASSPORT          36  ██
IMEI              33  ██

---

Dataset Structure

Each sample in the dataset is a JSON object with the following fields:

{
  "text": "Patiënt Fatima El-Amrani werd op 14 maart 2023 om 08:30 opgenomen in het LUMC te Leiden.",
  "entities": [
    {
      "entity_type": "PERSON",
      "start": 8,
      "end": 24,
      "text": "Fatima El-Amrani"
    },
    {
      "entity_type": "DATE",
      "start": 33,
      "end": 46,
      "text": "14 maart 2023"
    },
    {
      "entity_type": "TIME",
      "start": 50,
      "end": 55,
      "text": "08:30"
    },
    {
      "entity_type": "LOCATION",
      "start": 73,
      "end": 77,
      "text": "LUMC"
    },
    {
      "entity_type": "LOCATION",
      "start": 81,
      "end": 87,
      "text": "Leiden"
    }
  ]
}

• text — The raw Dutch text sample.
• entities — List of annotated entities, each with:
  • entity_type — One of the 19 entity types listed above.
  • start — Character-level start index (inclusive).
  • end — Character-level end index (exclusive).
  • text — The exact entity substring, satisfying text[start:end] == entity.text.

---

Intended Use

• Training and fine-tuning Dutch NER models for PII detection
• Developing and benchmarking text anonymization pipelines
• Testing data masking tools for Dutch healthcare documents
• Research on privacy-preserving NLP in Dutch medical contexts

---

Citation

If you use this dataset in your research or project, please cite it as:

@dataset{dutch_healthcare_pii_ner_2026,
  title     = {Dutch Healthcare PII NER Dataset},
  author    = {Careons},
  year      = {2026},
  note      = {Synthetic dataset generated with LLM assistance},
  url       = {https://huggingface.co/datasets/careons/dutch-healthcare-pii-ner}
}

---

Related Project

This dataset is part of Careons's open-source Dutch health informatics initiative. Visit our GitHub organization for the full anonymization pipeline, model checkpoints, and tooling built on top of this data.