DedalusHealthCare/tinybert-ner-demo-en

Demo - EN

Model Description

Named Entity Recognition model for demo in EN.

• Base Model: DedalusHealthCare/tinybert-mlm-en
• Architecture: tinybert

Intended Use

This model is designed for medical and clinical NLP tasks. It should be used by healthcare professionals and researchers for:

• Clinical text analysis
• Medical document processing
• Healthcare data extraction

Not intended for:

• Direct clinical decision making without human oversight
• Use outside the trained domain or language

Evaluation Results

Evaluation Set: goldset

Overall Performance

Metric	Score
Precision	0.0221
Recall	0.1395
F1	0.0382
Accuracy	0.1522

Per-Class Performance

Class	Precision	Recall	F1-Score	Support
DISORDER_FINDING	0.0221	0.1395	0.0382	86

Usage Example

PyTorch

from transformers import AutoTokenizer, AutoModelForTokenClassification, pipeline

model_name = "DedalusHealthCare/tinybert-demo-en"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForTokenClassification.from_pretrained(model_name)

# Using pipeline
ner_pipeline = pipeline("token-classification", model=model, tokenizer=tokenizer, aggregation_strategy="simple")
text = "Patient has diabetes and hypertension."
entities = ner_pipeline(text)

for entity in entities:
    print(f"{entity['entity_group']}: {entity['word']} ({entity['score']:.2f})")

ONNX Runtime (recommended for inference)

from optimum.onnxruntime import ORTModelForTokenClassification
from transformers import AutoTokenizer, pipeline

model_name = "DedalusHealthCare/tinybert-demo-en"
tokenizer = AutoTokenizer.from_pretrained(model_name)
onnx_model = ORTModelForTokenClassification.from_pretrained(model_name, subfolder="onnx")

ner_pipeline = pipeline("token-classification", model=onnx_model, tokenizer=tokenizer, aggregation_strategy="simple")
text = "Patient has diabetes and hypertension."
entities = ner_pipeline(text)

for entity in entities:
    print(f"{entity['entity_group']}: {entity['word']} ({entity['score']:.2f})")

Limitations

• Model is trained on medical/clinical text and may not generalize to other domains
• Performance may vary on text from different healthcare systems or documentation styles
• Should not be used as the sole basis for clinical decisions
• Requires validation in specific deployment contexts

Citation

If you use this model, please cite:

@misc{dedalus-medical-nlp,
  author = {DH Healthcare GmbH},
  title = {Medical NLP Models},
  year = {2025},
  publisher = {HuggingFace},
  howpublished = {\url{https://huggingface.co/DedalusHealthCare}}
}