adamleeit/biolink-lg-interop-ner-model

Healthcare Interoperability NER Model

This model is fine-tuned to identify healthcare interoperability concepts in biomedical text.

Model Details

• Base model: BioLinkBERT-large with domain adaptation on healthcare interoperability literature
• Task: Named Entity Recognition (NER)
• Training data: Annotated corpus of healthcare interoperability concepts
• Labels:
  • O: Not an entity
  • B-ENTITY: Beginning of interoperability entity
  • I-ENTITY: Inside/continuation of interoperability entity

Evaluation Results

Our model achieves the following performance on healthcare interoperability entity recognition:

Metric	Score
F1 Score	74.4%
Precision	75.4%
Recall	73.3%
Accuracy	94.3%

Improvement over base model:

Model	F1 Score	Precision	Recall
Base BioLinkBERT	65.9%	66.7%	65.2%
+ Domain Adaptation	74.4%	75.4%	73.3%
Improvement	+8.5%	+8.7%	+8.1%

Example model outputs:

Input: "FHIR and HL7 standards improve healthcare data interoperability."

Identified entities:

• FHIR
• HL7 standards
• healthcare data interoperability

Usage

from transformers import AutoTokenizer, AutoModelForTokenClassification
import torch
# Load model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("adamleeit/biolink-lg-interop-ner-model")
model = AutoModelForTokenClassification.from_pretrained("adamleeit/biolink-lg-interop-ner-model")
# Tokenize input text
text = "FHIR and HL7 standards improve healthcare data interoperability."
inputs = tokenizer(text, return_tensors="pt", return_offsets_mapping=True)
offset_mapping = inputs.pop("offset_mapping")
# Get predictions
with torch.no_grad():
    outputs = model(**inputs)
# Process predictions
predictions = torch.argmax(outputs.logits, dim=2)
input_ids = inputs["input_ids"][0]
# Convert predictions to entities
predicted_entities = []
current_entity = []
for i, pred in enumerate(predictions[0]):
    token = tokenizer.convert_ids_to_tokens(input_ids[i])
    if pred == 0:  # O
        if current_entity:
            predicted_entities.append(" ".join(current_entity))
            current_entity = []
    elif pred == 1:  # B-ENTITY
        if current_entity:
            predicted_entities.append(" ".join(current_entity))
        current_entity = [token]
    elif pred == 2:  # I-ENTITY
        current_entity.append(token)
if current_entity:
    predicted_entities.append(" ".join(current_entity))
print(f"Identified entities: {predicted_entities}")

Citation If you use this model in your research, please cite:

@misc{interoperability-ner-model,
  author = {Your Name},
  title = {Healthcare Interoperability NER Model},
  year = {2025},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/adamleeit/biolink-lg-interop-ner-model}}
}