http://doi.org/10.26347/1607-2502202309-10017-024
This text provides an overview of the importance of mapping Electronic Health Records (EHR) to the Human Phenotype Ontology (HPO) using advanced language models. It discusses the challenges posed by the complex and diverse language used in EHRs, which hinders information sharing and extraction. The paper explores the role of numerical and text data within EHRs and their significance in providing valuable insights into patient health. Current approaches using advanced language models like the Transformer, BERT (Bidirectional Encoder Representations from Transformer), BioBERT, ClinicalBERT, and BioALBERT are highlighted, showcasing their ability to capture complex dependencies and context in EHR data. The applications of these models in tasks such as named entity recognition, relation extraction, clinical concept normalization, question answering, and document classification are explored. The paper concludes by emphasizing the potential of advanced language models to improve clinical decision-making, research, and patient care by enabling accurate mapping of EHR data to standardized ontologies like HPO.