| Property | Value |
|---|---|
| License | Apache 2.0 |
| Language | Spanish |
| Paper | View Paper |
| Training Data | 1B+ tokens |
What is roberta-base-biomedical-clinical-es?
This is a specialized RoBERTa-based language model designed specifically for Spanish biomedical and clinical text processing. Developed by the Text Mining Unit at Barcelona Supercomputing Center, it has been trained on a massive corpus of over 1 billion tokens from various medical sources, including clinical documents, research papers, and medical crawl data.
Implementation Details
The model utilizes a byte-level BPE tokenizer with a 52,000 token vocabulary. Training was conducted using 16 NVIDIA V100 GPUs over 48 hours, implementing the masked language modeling approach with Adam optimizer and a 0.0005 peak learning rate. The training corpus combines cleaned biomedical texts with uncleaned clinical notes to maintain authentic medical language patterns.
- Trained on diverse medical sources including hospital discharge reports, clinical cases, and scientific publications
- Implements masked language modeling for fill-mask tasks
- Achieves state-of-the-art performance on Spanish medical NER tasks
Core Capabilities
- Masked language modeling for medical text completion
- Named Entity Recognition for medical terms (90.04% F1 score on PharmaCoNER)
- Clinical text understanding and processing
- Biomedical document analysis
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its specialized focus on Spanish medical text and its comprehensive training data from real clinical documents. It outperforms general-purpose models like mBERT and BETO on medical NER tasks, showing significant improvements in understanding medical terminology and context.
Q: What are the recommended use cases?
The model is primarily designed for masked language modeling tasks in medical contexts but can be fine-tuned for various downstream tasks including named entity recognition, text classification, and medical document analysis. It's particularly suitable for processing Spanish clinical notes, medical research papers, and healthcare-related content.