license: apache-2.0
language:
- es
base_model:
- PlanTL-GOB-ES/roberta-base-biomedical-clinical-es
tags:
- medical
- spanish
- bi-encoder
- entity-linking
- sapbert
- umls
- snomed-ct
MedProcNER-bi-encoder
Model Description
MedProcNER-bi-encoder is a domain-specific bi-encoder model for medical entity linking in Spanish, trained using synonym pairs from the MedProcNER corpus and SNOMED-CT (Fully Specified Name and preferred synonyms). The training data was curated from the gold standard corpus and enriched with knowledge-based synonyms to enhance entity normalization tasks.
💡 Intended Use
- Domain: Spanish Clinical NLP
- Tasks: Entity linking of MedProcNER mentions to SNOMED-CT concepts
- Evaluated On: MedProcNER (Gold Standard, Unseen Mentions, Unseen Codes)
- Users: Researchers and developers focusing on specialized medical NEL
💬 Definitions
- Unseen Mentions: Mentions that do not appear in training but reference known codes.
- Unseen Codes: Mentions associated with SNOMED-CT codes never seen during training.
📈 Performance Summary (Top-25 Accuracy)
| Evaluation Split | Top-25 Accuracy |
|---|---|
| Gold Standard | 0.917 |
| Unseen Mentions | 0.831 |
| Unseen Codes | 0.808 |
🧪 Usage
from transformers import AutoModel, AutoTokenizer
import torch
model = AutoModel.from_pretrained("ICB-UMA/MedProcNER-bi-encoder")
tokenizer = AutoTokenizer.from_pretrained("ICB-UMA/MedProcNER-bi-encoder")
mention = "insuficiencia renal aguda"
inputs = tokenizer(mention, return_tensors="pt")
with torch.no_grad():
outputs = model(**inputs)
embedding = outputs.last_hidden_state[:, 0, :]
print(embedding.shape)
Use with Faiss or FaissEncoder for efficient retrieval.
⚠️ Limitations
- The model is specialized for MedProcNER mentions and may underperform in other domains or corpora.
- Expert supervision is advised for clinical deployment.
📚 Citation
Gallego, Fernando and López-García, Guillermo and Gasco, Luis and Krallinger, Martin and Veredas, Francisco J., Clinlinker-Kb: Clinical Entity Linking in Spanish with Knowledge-Graph Enhanced Biencoders. Available at SSRN: http://dx.doi.org/10.2139/ssrn.4939986
Authors
Fernando Gallego, Guillermo López-García, Luis Gasco-Sánchez, Martin Krallinger, Francisco J Veredas