event-situation-es-cat-tagger-v1
Table of contents
Click to expand
Model description
A fine-tuned version of the
bsc-bio-ehr-es model for the recognition of clinical events and situations in Spanish and Catalan clinical text and electronic health records. The model labels the following entity classes:
| Label | Description |
|---|
EVENTO | Clinical events — occurrences with a defined onset (e.g. symptoms, procedures, diagnoses as they happen). |
SITUACION | Clinical situations / states — ongoing conditions, contexts or background circumstances. |
Labels follow the BIO tagging scheme (B-, I-, O).
Training data
The model was trained on a corpus of 200 pediatric clinical histories manually annotated for events and situations:
- ⅓ Catalan (66 documents)
- ⅔ Spanish (133 documents)
- 50 documents are from psychiatry, which contain a higher density of event and situation mentions.
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 | 10 |
| Batch size | {8, 16} |
| Learning rate | log-uniform, 1e-7 – 1e-3 |
| Classifier dropout | uniform, 0.2 – 0.8 (step 0.2) |
| Weight decay | {0.01, 0.1} |
| Warmup ratio | 0.1 |
| Class weight strategy | {none, freq_sqrt} |
| 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
F-score on the test set under two regimes — strict (correct class and exact mention boundaries) and overlapping (correct class with some boundary overlap, exact span not required):
| Label | Strict F-Score | Overlapping F-Score |
|---|
EVENTO | 0.190 | 0.444 |
SITUACION | 0.336 | 0.563 |
| micro avg | 0.307 | 0.539 |
These are early results from a single model trained on a small corpus; strict boundary matching is notably harder than overlap-based matching for these entity types.
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.
