GERNERMED - An Open German Medical NER Model

About

The is the project repository for GERNERMED, a named entity recognition (NER) model in the context of German medical natural language processing (NLP).
In particular, GERNERMED is the first open neural NER model for medical entities designed for German data.

Our follow-up work, GERNERMED++, is available here: https://github.com/frankkramer-lab/GERNERMED-pp

Published papers:
See our short, software-related paper at: https://doi.org/10.1016/j.simpa.2021.100212
Or see our long paper at: https://doi.org/10.2196/39077

NER entities: The following entities are supported:

• Drug
• Strength
• Route
• Form
• Dosage
• Frequency
• Duration

The evaluation scores on the test set are as follows:

NER Tag	Precision	Recall	F1-Score
Drug	67.33	66.17	66.74
Strength	92.34	90.99	91.66
Route	89.93	90.14	90.04
Form	91.94	89.24	90.57
Dosage	87.83	87.57	87.70
Frequency	79.14	76.92	78.01
Duration	67.86	52.78	59.37
total	82.31	80.79	81.54

NER Demonstration

Setup and Usage

For a quick test, run python3 example.py to run the NER model pipeline.

The package NER pipeline is available in /data/de_GERNERMED-1.0.0.tar.gz and can be installed by:

# From local fs
python3 -m pip install ./data/de_GERNERMED-1.0.0.tar.gz

# From GitHub
python3 -m pip install https://github.com/frankkramer-lab/GERNERMED/blob/main/data/de_GERNERMED-1.0.0.tar.gz?raw=true

The pipeline can be used in Python:

import spacy

nlp = spacy.load("de_GERNERMED")
doc = nlp("Dem Patienten wurde die Einnahme von Paracetamol (500 mg, zwei Tabletten täglich, 8 Wochen lang) zur Behandlung empfohlen.")

# Show entities
print(doc.ents)

Reproducibility

The evaluation scores on the testset can be obtained by ./run_eval.sh.
The custom German dataset is stored in /data/GERNERMED_dataset.json.

Citation

Long-form JMIR Formative Research paper (https://doi.org/10.2196/39077):

@article{info:doi/10.2196/39077,
  author="Frei, Johann and Kramer, Frank",
  title="German Medical Named Entity Recognition Model and Data Set Creation Using Machine Translation and Word Alignment: Algorithm Development and Validation",
  journal="JMIR Form Res",
  year="2023",
  month="Feb",
  day="28",
  volume="7",
  pages="e39077",
  keywords="natural language processing; named entity recognition; information extraction",
  issn="2561-326X",
  doi="10.2196/39077",
  url="https://formative.jmir.org/2023/1/e39077",
  url="https://doi.org/10.2196/39077",
  url="http://www.ncbi.nlm.nih.gov/pubmed/36853741"
}

Short, software-focused paper at Software Impacts (https://doi.org/10.1016/j.simpa.2021.100212):

@article{frei_gernermed_2022,
  title = {{GERNERMED}: An open German medical {NER} model},
  volume = {11},
  issn = {2665-9638},
  url = {https://www.sciencedirect.com/science/article/pii/S2665963821000944},
  doi = {10.1016/j.simpa.2021.100212},
  pages = {100212},
  journaltitle = {Software Impacts},
  shortjournal = {Software Impacts},
  author = {Frei, Johann and Kramer, Frank},
  urldate = {2022-02-21},
  date = {2022-02-01},
  langid = {english},
  keywords = {Machine learning, Natural language processing, Clinical text mining, Named entity recognition},
}

The ArXiv pre-print paper from http://arxiv.org/abs/2109.12104

@misc{frei2021gernermed,
  title={GERNERMED -- An Open German Medical NER Model}, 
  author={Johann Frei and Frank Kramer},
  year={2021},
  eprint={2109.12104},
  archivePrefix={arXiv},
  primaryClass={cs.CL}
}