VICOMTECH at MedProcNER 2023: Transformers-based Sequence-labelling and Cross-encoding for Entity Detection and Normalisation in Spanish Clinical Texts

@Article {
title = {VICOMTECH at MedProcNER 2023: Transformers-based Sequence-labelling and Cross-encoding for Entity Detection and Normalisation in Spanish Clinical Texts},
pages = {206-218},
keywds = {
Clinical Coding; Document Indexing; Entity Linking; Entity Normalisation; Named Entity Recognition; SNOMED CT
}
abstract = {

This paper describes the participation of the Vicomtech NLP team in the MedProcNER 2023 shared task about detecting mentions of procedures in clinical texts written in Spanish and normalising them to SNOMED CT codes. We participate in each of the three tasks, combining multiple approaches and strategies. For Task 1 (NER) we use a Transformer-based model to perform sequence labelling. For Task 2 (Normalisation) we use Semantic Text Search approaches to relate entity mentions to their codes. The solution for Task 3 (Indexing) is built on top of the two first tasks. For Task 1 our system obtained 77.96% of F1-score. Our approaches for Task 2 and Task 3 achieved the highest F1 scores in the official evaluation results—57.07% and 62.42%, respectively.

}
date = {2023-09-18},
}