EVALUATION OF SPACY AND DEEPPAVLOV LIBRARY TOOLS FOR NAMED ENTITIES RECOGNITION FROM DESCRIPTIONS OF EXAMINATION RESULTS OF PATIENTS WITH COVID-19

Relevance. Determined by the need to extract significant features from electronic medical records to automate the assessment of patients' condition. Aim. Assessing the possibility of identifying named entitie in electronic descriptions of examinations of patients with COVID-19 using the BERT mo...

Full description

Saved in:
Bibliographic Details
Main Authors: Dmitry E. Sokolovsky, Vladimir N. Nekrasov, Sergey A. Zemlyansky, Sergey V. Axyonov
Format: Article
Language:English
Published: Tomsk Polytechnic University 2023-06-01
Series:Известия Томского политехнического университета: Промышленная кибернетика
Subjects:
Online Access:https://indcyb.ru/journal/article/view/27/22
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849433514437509120
author Dmitry E. Sokolovsky
Vladimir N. Nekrasov
Sergey A. Zemlyansky
Sergey V. Axyonov
author_facet Dmitry E. Sokolovsky
Vladimir N. Nekrasov
Sergey A. Zemlyansky
Sergey V. Axyonov
author_sort Dmitry E. Sokolovsky
collection DOAJ
description Relevance. Determined by the need to extract significant features from electronic medical records to automate the assessment of patients' condition. Aim. Assessing the possibility of identifying named entitie in electronic descriptions of examinations of patients with COVID-19 using the BERT model from the SpaCy and DeepPavlov libraries.Methods. De ep learning, statistical methods. Results and conclusions. The authors have carried out a fine-tuning study on BERT neural network models from the SpaCy and DeepPavlov libraries to annotate documents “Examination of patients by the attending physician” in order to highlight the following predictors of patient assessment: temperature, blood pressure, respiratory rate, heart rate and saturation. Configuration and evaluation of the effectiveness of the architectures was carried out based on the markup of 340 anonymized electronic medical records of patients with COVID-19, obtained using the SibMED Data Clinical Repository service. It is shown that setting up models on a number of about 150 labeled documents makes it possible to determine the specified predictors in such texts with accuracy (Precision) of 85–98% and completeness (Recall) of 77–98%, depending on the predictor. The quality metrics of the architectures from the selected libraries differed slightly. Iterative expansion of the training set as a result of the operation of models with subsequent additional tuning leads to an increase in the effectiveness of the models.
format Article
id doaj-art-9b2c074e73524213885e1241f942252e
institution Kabale University
issn 2949-5407
language English
publishDate 2023-06-01
publisher Tomsk Polytechnic University
record_format Article
series Известия Томского политехнического университета: Промышленная кибернетика
spelling doaj-art-9b2c074e73524213885e1241f942252e2025-08-20T03:27:01ZengTomsk Polytechnic UniversityИзвестия Томского политехнического университета: Промышленная кибернетика2949-54072023-06-0112465310.18799/29495407/2023/2/27EVALUATION OF SPACY AND DEEPPAVLOV LIBRARY TOOLS FOR NAMED ENTITIES RECOGNITION FROM DESCRIPTIONS OF EXAMINATION RESULTS OF PATIENTS WITH COVID-19Dmitry E. Sokolovsky0Vladimir N. Nekrasov1Sergey A. Zemlyansky2Sergey V. Axyonov3National Research Tomsk Polytechnic University, RussiaS.M. Kirov Military Medical Academy, Russia National Research Tomsk State University, RussiaNational Research Tomsk Polytechnic University, RussiaRelevance. Determined by the need to extract significant features from electronic medical records to automate the assessment of patients' condition. Aim. Assessing the possibility of identifying named entitie in electronic descriptions of examinations of patients with COVID-19 using the BERT model from the SpaCy and DeepPavlov libraries.Methods. De ep learning, statistical methods. Results and conclusions. The authors have carried out a fine-tuning study on BERT neural network models from the SpaCy and DeepPavlov libraries to annotate documents “Examination of patients by the attending physician” in order to highlight the following predictors of patient assessment: temperature, blood pressure, respiratory rate, heart rate and saturation. Configuration and evaluation of the effectiveness of the architectures was carried out based on the markup of 340 anonymized electronic medical records of patients with COVID-19, obtained using the SibMED Data Clinical Repository service. It is shown that setting up models on a number of about 150 labeled documents makes it possible to determine the specified predictors in such texts with accuracy (Precision) of 85–98% and completeness (Recall) of 77–98%, depending on the predictor. The quality metrics of the architectures from the selected libraries differed slightly. Iterative expansion of the training set as a result of the operation of models with subsequent additional tuning leads to an increase in the effectiveness of the models.https://indcyb.ru/journal/article/view/27/22deep learningnamed entity extractionbertspaсydeeppavlov
spellingShingle Dmitry E. Sokolovsky
Vladimir N. Nekrasov
Sergey A. Zemlyansky
Sergey V. Axyonov
EVALUATION OF SPACY AND DEEPPAVLOV LIBRARY TOOLS FOR NAMED ENTITIES RECOGNITION FROM DESCRIPTIONS OF EXAMINATION RESULTS OF PATIENTS WITH COVID-19
Известия Томского политехнического университета: Промышленная кибернетика
deep learning
named entity extraction
bert
spaсy
deeppavlov
title EVALUATION OF SPACY AND DEEPPAVLOV LIBRARY TOOLS FOR NAMED ENTITIES RECOGNITION FROM DESCRIPTIONS OF EXAMINATION RESULTS OF PATIENTS WITH COVID-19
title_full EVALUATION OF SPACY AND DEEPPAVLOV LIBRARY TOOLS FOR NAMED ENTITIES RECOGNITION FROM DESCRIPTIONS OF EXAMINATION RESULTS OF PATIENTS WITH COVID-19
title_fullStr EVALUATION OF SPACY AND DEEPPAVLOV LIBRARY TOOLS FOR NAMED ENTITIES RECOGNITION FROM DESCRIPTIONS OF EXAMINATION RESULTS OF PATIENTS WITH COVID-19
title_full_unstemmed EVALUATION OF SPACY AND DEEPPAVLOV LIBRARY TOOLS FOR NAMED ENTITIES RECOGNITION FROM DESCRIPTIONS OF EXAMINATION RESULTS OF PATIENTS WITH COVID-19
title_short EVALUATION OF SPACY AND DEEPPAVLOV LIBRARY TOOLS FOR NAMED ENTITIES RECOGNITION FROM DESCRIPTIONS OF EXAMINATION RESULTS OF PATIENTS WITH COVID-19
title_sort evaluation of spacy and deeppavlov library tools for named entities recognition from descriptions of examination results of patients with covid 19
topic deep learning
named entity extraction
bert
spaсy
deeppavlov
url https://indcyb.ru/journal/article/view/27/22
work_keys_str_mv AT dmitryesokolovsky evaluationofspacyanddeeppavlovlibrarytoolsfornamedentitiesrecognitionfromdescriptionsofexaminationresultsofpatientswithcovid19
AT vladimirnnekrasov evaluationofspacyanddeeppavlovlibrarytoolsfornamedentitiesrecognitionfromdescriptionsofexaminationresultsofpatientswithcovid19
AT sergeyazemlyansky evaluationofspacyanddeeppavlovlibrarytoolsfornamedentitiesrecognitionfromdescriptionsofexaminationresultsofpatientswithcovid19
AT sergeyvaxyonov evaluationofspacyanddeeppavlovlibrarytoolsfornamedentitiesrecognitionfromdescriptionsofexaminationresultsofpatientswithcovid19