Public artifacts¶

The MedDeID Hugging Face collection contains the public Dutch and English models, development data, and independent benchmarks.

Choose what you need¶

Goal	Artifact
Try MedDeID without installing it	Non-clinical demo
De-identify Dutch clinical text	Dutch synthetic model
De-identify English clinical text	English synthetic model
Develop or adapt a model	A synthetic corpus for the applicable language
Evaluate a system independently	The applicable synthetic benchmark
Prepare human annotations	The Dutch or English annotation guideline

The published models are synthetic baselines. Before using one with clinical text, validate it on representative data from your own setting.

Run a published model¶

`stighellemans/meddeid-dutch-synth`¶

Choose this model for Dutch text. MedDeID downloads it on first use and then reuses the local Hugging Face cache:

pip install meddeid
meddeid deidentify note.txt \
  --model stighellemans/meddeid-dutch-synth

`stighellemans/meddeid-english-synth`¶

Choose this model for English text and select the regional profile that matches the document formats:

meddeid deidentify note.txt \
  --model stighellemans/meddeid-english-synth \
  --language-profile en-GB  # Use en-US for US formats

For a reproducible workflow, record the model's immutable Hub revision. To run from a local copy, keep the complete model directory together.

Develop or evaluate with published data¶

Use a corpus for model development, including training and validation. Keep the benchmark separate until final evaluation so it remains an independent test.

Language	Development corpus	Independent benchmark
Dutch	`meddeid-dutch-synthetic-corpus`	`meddeid-dutch-synthetic-benchmark`
English	`meddeid-english-synthetic-corpus`	`meddeid-english-synthetic-benchmark`

Before downloading or citing an artifact, read its card for the exact contents, provenance, evaluation results, limitations, licence, and available revisions. Pin the immutable revision when the result must be reproducible.

Annotation guidelines¶

The current ProductionLabels_v1 guidelines are available as PDF reading copies and editable DOCX sources in the Zenodo releases for Dutch and English. They explain what annotators should mark; the precise software rules are documented separately in meddeid-core.

Public demo¶

The MedDeID hosted demo runs the public synthetic model for non-sensitive examples. It executes on Hugging Face infrastructure: never paste real patient or caregiver information. Use meddeid locally for clinical text.