bsc-bio-ehr-es-temporalidad

Table of contents

Click to expand

• Model description
• How to use
• Evaluation
• Limitations and bias
• Additional information
  • Authors
  • Contact information
  • Funding
  • Citing information
  • Disclaimer

Model description

A fine-tuned version of the bsc-bio-ehr-es model for the recognition of temporal expressions (TIMEX-style) in Spanish clinical text and electronic health records. The model labels the following entity classes:

Label	Description
AGE	Age expressions (e.g. 67 años).
DATE	Calendar dates and references to a point in time (e.g. en 2018).
DURATION	Durations / spans of time (e.g. 3 días de evolución).
SET	Recurring or periodic expressions (e.g. cada 4 horas, mensuales).
TIME	Times of day (e.g. por la noche).

Labels follow the BIO tagging scheme (B-, I-, O).

Training data

The model was trained on an internal Spanish clinical corpus (temporalidad-ner) annotated for temporal expressions following a TIMEX-style scheme.

Training procedure

The model was selected through a Weights & Biases hyperparameter sweep (up to 50 runs), keeping the configuration that maximized validation micro-F1 (eval/f1) over the bsc-bio-ehr-es base model. The search space explored was:

Hyperparameter	Search space
Base model	PlanTL-GOB-ES/bsc-bio-ehr-es
Epochs	15
Batch size	{8, 16, 32}
Learning rate	log-uniform, 1e-7 – 1e-3
Classifier dropout	uniform, 0.1 – 0.8 (step 0.1)
Weight decay	{0, 1e-5, 1e-4, 1e-3, 0.01, 0.1}
Warmup ratio	{0, 0.1}
Class weight strategy	none
Evaluation strategy	per epoch

How to use

⚠ We recommend pre-tokenizing the input text into words instead of providing it directly to the model, as this is how the model was trained. Otherwise, the results and performance might get affected.

A usage example can be found here.

Evaluation

Strict (same class, exact boundaries) and overlapping (same class, some boundary overlap) precision / recall / F-score on the test set.

Label	Precision	Recall	F-Score	Ov. Precision	Ov. Recall	Ov. F-Score
AGE	0.9426	0.9287	0.9356	0.9596	0.9455	0.9525
DATE	0.9093	0.9167	0.9130	0.9469	0.9546	0.9508
DURATION	0.9013	0.9129	0.9071	0.9384	0.9505	0.9444
SET	0.8466	0.9096	0.8770	0.8886	0.9548	0.9205
TIME	0.6323	0.7656	0.6926	0.7097	0.8594	0.7774
micro avg	0.8866	0.9107	0.8985	0.9241	0.9492	0.9365

Limitations and bias

At the time of submission, no measures have been taken to estimate the bias embedded in the model. However, we are well aware that our models may be biased since the corpora have been collected using crawling techniques on multiple web sources. We intend to conduct research in these areas in the future, and if completed, this model card will be updated.

Additional information

Authors

NLP4BIA team at the Barcelona Supercomputing Center (nlp4bia@bsc.es).

Contact information

jan.rodriguez [at] bsc.es

Funding

More information will be available soon.

Citing information

Please cite the following works:

Disclaimer

The models published in this repository are intended for a generalist purpose and are available to third parties. These models may have bias and/or any other undesirable distortions.

When third parties deploy or provide systems and/or services to other parties using any of these models (or using systems based on these models) or become users of the models, they should note that it is their responsibility to mitigate the risks arising from their use and, in any event, to comply with applicable regulations, including regulations regarding the use of artificial intelligence.

---

Los modelos publicados en este repositorio tienen una finalidad generalista y están a disposición de terceros. Estos modelos pueden tener sesgos y/u otro tipo de distorsiones indeseables.

Cuando terceros desplieguen o proporcionen sistemas y/o servicios a otras partes usando alguno de estos modelos (o utilizando sistemas basados en estos modelos) o se conviertan en usuarios de los modelos, deben tener en cuenta que es su responsabilidad mitigar los riesgos derivados de su uso y, en todo caso, cumplir con la normativa aplicable, incluyendo la normativa en materia de uso de inteligencia artificial.