Negation And Uncertainty Detection in clinical texts written in Spanish: A deep learning-based approach
This repository contains a deep learning-based approach for uncertainty and negation detection from clinical texts written in Spanish.
The approach explores two deep learning methods to perform negation and uncertainy detection from clinical text written in Spanish: BILSTM and BERT:- Bidirectional Long Short memory (BiLSTM-CRF): This method consist of three layers: Embedding layer, BiLSTM layer, and CRF layer. The directory "BiLSTM" contains the implementation for this method.
- Embeddings We used two types of embeddings: biomedical embeddings and clinical embeddings. Biomedical embeddings for the Spanish language [2] can be download from Zenodo. Clinical embeddings were trained with more than 1 million of clinical notes of two public hospitals in Spain and Colombia. Clinical embeddings can be available only after an evaluation of the hospital's ethics committee
- Bidirectional Encoder Representation for Transformers (BERT): We use the pre-trained BERT model fine tune with a classification layer on top. We use Multilingual BERT as contextualized embeddings. The directory "BERT" contains the implementation for this method.
- The NUBES corpus is available from https://github.com/Vicomtech/NUBes-negation-uncertainty-biomedical-corpus
- The IULA corpus is publicly available and can be access from http://eines.iula.upf.edu/brat/#/NegationOnCR_IULA
- The cancer dataset is an in-house manually annotated corpus with data from patients treated with lung and breast cancer. This dataset was used only for validation purposes. This dataset is affected by General Data Protection Regulation (GDPR) and is available only "upon request". This dataset can be accessible after an evaluation by the hospital's ethics committee.
References:
- Lample, G.; Ballesteros, M.; Subramanian, S.; Kawakami, K.; Dyer, C. Neural architectures for named entity recognition.2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL HLT 2016
- Soares, F.; Villegas, M.; Gonzalez-Agirre, A.; Krallinger, M.; Armengol-Estapé, J. Medical Word Embeddings787for Spanish: Development and Evaluation. Proceedings of the 2nd Clinical Natural Language Processing788Workshop; Association for Computational Linguistics: Minneapolis, Minnesota, USA, 2019
- Lima, S.; Perez, N.; Cuadros, M.; Rigau, G. NUBES: A Corpus of Negation and Uncertainty in Spanish Clinical Texts 2020. Proceedings of the Workshop Computational Semantics Beyond Events and Roles, Valencia, Spain.
- Marimon, M.; Vivaldi, J.; Bel, N. Annotation of negation in the IULA Spanish Clinical Record Corpus. Proceedings of the Workshop Computational Semantics
Beyond Events and Roles; Association for807Computational Linguistics: Valencia, Spain, 2017; pp. 43–52. doi:10.18653/v1/W17-1807
- Devlin, J.; Chang, M.W.; Lee, K.; Toutanova, K. BERT: Pre-training of deep bidirectional transformers for795language understanding.NAACL HLT 2019 - 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies - Proceedings of the Conference 2019