⚠️NOTICE⚠️: THIS MODEL HAS BEEN MOVED TO THE FOLLOWING URL AND WILL SOON BE REMOVED: https://huggingface.co/PlanTL-GOB-ES/roberta-base-biomedical-clinical-es

Biomedical-clinical language model for Spanish

Biomedical pretrained language model for Spanish. For more details about the corpus, the pretraining and the evaluation, check the official repository and read our preprint "Carrino, C. P., Armengol-Estapé, J., Gutiérrez-Fandiño, A., Llop-Palao, J., Pàmies, M., Gonzalez-Agirre, A., & Villegas, M. (2021). Biomedical and Clinical Language Models for Spanish: On the Benefits of Domain-Specific Pretraining in a Mid-Resource Scenario.".

Tokenization and model pretraining

This model is a RoBERTa-based model trained on a biomedical-clinical corpus in Spanish collected from several sources (see next section). The training corpus has been tokenized using a byte version of Byte-Pair Encoding (BPE) used in the original RoBERTA model with a vocabulary size of 52,000 tokens. The pretraining consists of a masked language model training at the subword level following the approach employed for the RoBERTa base model with the same hyperparameters as in the original work. The training lasted a total of 48 hours with 16 NVIDIA V100 GPUs of 16GB DDRAM, using Adam optimizer with a peak learning rate of 0.0005 and an effective batch size of 2,048 sentences.

Training corpora and preprocessing

The training corpus is composed of several biomedical corpora in Spanish, collected from publicly available corpora and crawlers, and a real-world clinical corpus collected from more than 278K clinical documents and notes. To obtain a high-quality training corpus while retaining the idiosyncrasies of the clinical language, a cleaning pipeline has been applied only to the biomedical corpora, keeping the clinical corpus uncleaned. Essentially, the cleaning operations used are:

• data parsing in different formats
  • sentence splitting
  • language detection
  • filtering of ill-formed sentences
  • deduplication of repetitive contents
  • keep the original document boundaries

Then, the biomedical corpora are concatenated and further global deduplication among the biomedical corpora have been applied. Eventually, the clinical corpus is concatenated to the cleaned biomedical corpus resulting in a medium-size biomedical-clinical corpus for Spanish composed of more than 1B tokens. The table below shows some basic statistics of the individual cleaned corpora:

Evaluation and results

The model has been evaluated on the Named Entity Recognition (NER) using the following datasets:

• PharmaCoNER: is a track on chemical and drug mention recognition from Spanish medical texts (for more info see: https://temu.bsc.es/pharmaconer/).

• CANTEMIST: is a shared task specifically focusing on named entity recognition of tumor morphology, in Spanish (for more info see: https://zenodo.org/record/3978041#.YTt5qH2xXbQ).

• ICTUSnet: consists of 1,006 hospital discharge reports of patients admitted for stroke from 18 different Spanish hospitals. It contains more than 79,000 annotations for 51 different kinds of variables.

The evaluation results are compared against the mBERT and BETO models:

Intended uses & limitations

The model is ready-to-use only for masked language modelling to perform the Fill Mask task (try the inference API or read the next section)

However, the is intended to be fine-tuned on downstream tasks such as Named Entity Recognition or Text Classification.

Cite

If you use our models, please cite our latest preprint:

@misc{carrino2021biomedical,
      title={Biomedical and Clinical Language Models for Spanish: On the Benefits of Domain-Specific Pretraining in a Mid-Resource Scenario}, 
      author={Casimiro Pio Carrino and Jordi Armengol-Estapé and Asier Gutiérrez-Fandiño and Joan Llop-Palao and Marc Pàmies and Aitor Gonzalez-Agirre and Marta Villegas},
      year={2021},
      eprint={2109.03570},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

If you use our Medical Crawler corpus, please cite the preprint:

@misc{carrino2021spanish,
      title={Spanish Biomedical Crawled Corpus: A Large, Diverse Dataset for Spanish Biomedical Language Models}, 
      author={Casimiro Pio Carrino and Jordi Armengol-Estapé and Ona de Gibert Bonet and Asier Gutiérrez-Fandiño and Aitor Gonzalez-Agirre and Martin Krallinger and Marta Villegas},
      year={2021},
      eprint={2109.07765},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

How to use

from transformers import AutoTokenizer, AutoModelForMaskedLM

tokenizer = AutoTokenizer.from_pretrained("BSC-TeMU/roberta-base-biomedical-es")

model = AutoModelForMaskedLM.from_pretrained("BSC-TeMU/roberta-base-biomedical-es")

from transformers import pipeline

unmasker = pipeline('fill-mask', model="BSC-TeMU/roberta-base-biomedical-es")

unmasker("El único antecedente personal a reseñar era la <mask> arterial.")

# Output
[
  {
    "sequence": " El único antecedente personal a reseñar era la hipertensión arterial.",
    "score": 0.9855039715766907,
    "token": 3529,
    "token_str": " hipertensión"
  },
  {
    "sequence": " El único antecedente personal a reseñar era la diabetes arterial.",
    "score": 0.0039140828885138035,
    "token": 1945,
    "token_str": " diabetes"
  },
  {
    "sequence": " El único antecedente personal a reseñar era la hipotensión arterial.",
    "score": 0.002484665485098958,
    "token": 11483,
    "token_str": " hipotensión"
  },
  {
    "sequence": " El único antecedente personal a reseñar era la Hipertensión arterial.",
    "score": 0.0023484621196985245,
    "token": 12238,
    "token_str": " Hipertensión"
  },
  {
    "sequence": " El único antecedente personal a reseñar era la presión arterial.",
    "score": 0.0008009297889657319,
    "token": 2267,
    "token_str": " presión"
  }
]