metadata
library_name: transformers
tags:
- contrastive-learning
- Spanish-UMLS
- Hierarchical-enrichment
- entity-linking
- biomedical
- spanish
license: mit
language:
- es
base_model:
- PlanTL-GOB-ES/roberta-base-biomedical-clinical-es
HERBERT: Leveraging UMLS Hierarchical Knowledge to Enhance Clinical Entity Normalization in Spanish
HERBERT-P is a contrastive-learning-based bi-encoder for medical entity normalization in Spanish, leveraging synonym and parent relationships from UMLS to enhance candidate retrieval for entity linking in clinical texts.
Key features:
- Base model: PlanTL-GOB-ES/roberta-base-biomedical-clinical-es
- Trained with 30 positive pairs per anchor (synonyms + parents)
- Task: Normalization of disease, procedure, and symptom mentions to SNOMED-CT/UMLS codes.
- Domain: Spanish biomedical/clinical texts.
- Corpora: DisTEMIST, MedProcNER, SympTEMIST.
Benchmark Results
| Corpus | Top-1 | Top-5 | Top-25 | Top-200 |
|---|---|---|---|---|
| DisTEMIST | 0.588 | 0.723 | 0.803 | 0.867 |
| SympTEMIST | 0.635 | 0.784 | 0.882 | 0.946 |
| MedProcNER | 0.651 | 0.765 | 0.838 | 0.892 |