Clinical Flair: A Pre-Trained Language Model for Spanish Clinical Natural Language Processing

• DOI: 10.18653/v1/2022.clinicalnlp-1.9
• Publication Date: 2022-07-26T02:59:46Z
• Authors: Matías Rojas; Jocelyn Dunstan; Fabián Villena
• Keywords: Sequence labeling

Word embeddings have been widely used in Natural Language Processing (NLP) tasks. Although these representations can capture the semantic information of words, they cannot learn sequence-level semantics. This problem be handled using contextual word derived from pre-trained language models, which contributed to significant improvements several NLP Further are achieved when pre-training models on domain-specific corpora. In this paper, we introduce Clinical Flair, a model trained Spanish clinical narratives. To validate quality retrieved our model, tested them four named entity recognition datasets belonging and biomedical domains. Our experiments confirm that incorporating into classical sequence labeling architectures improves performance dramatically compared general-domain embeddings, demonstrating importance having resources available.