An Accuracy Comparison of the Joint and Sequential Approaches for End-to-End Related Named Entities Extraction in the Texts of Russian-Language Reviews Based on Neural Networks

Sboev Alexander, Sboev Alexander, Roman Rybka, Roman Rybka, Aleksandr Naumov, Anton Selivanov, Artem Gryaznov, Ivan Moloshnikov

2022

Abstract

Solving a problem of relations recognition among significant pharmacological entities is one of the important stage of complex automatic analysis of drug reviews for purposes of pharmacovigilance, marketing, social situation analysis, healthcare, and others. The closest statement of the problem to practical cases is an end-to-end model to extract related entities from the scratch, with realization of two stages: recognition of significant entities (NER) and extraction of relation between them (RE). To our knowledge, this problem has not been solved for Russian drug review texts of every day lexis. So, there is no evaluation of the accuracy of its solution. A creation of the Russian Drug Review Corpus RDRS allowed to obtain such an evaluation presented in this work. We use two models for this purpose: the first is of joint NER and RE extraction, the second is of the step by step calculations, initially of NER and then RE. The difference in results, obtained on the basis of the above two ways, was analyzed. Both approaches demonstrated the close average accuracies of end-to-end solution, establishing an accuracy level of the problem in view about 51% f1 for the set of related entities: ADR-Drugname, Drugname- Diseasename, Drugname- SourceInfoDrug, Diseasename- Indication.

Paper Citation

in Harvard Style

Alexander S., Rybka R., Naumov A., Selivanov A., Gryaznov A. and Moloshnikov I. (2022). An Accuracy Comparison of the Joint and Sequential Approaches for End-to-End Related Named Entities Extraction in the Texts of Russian-Language Reviews Based on Neural Networks. In Proceedings of the 3rd International Symposium on Automation, Information and Computing - Volume 1: ISAIC; ISBN 978-989-758-622-4, SciTePress, pages 348-353. DOI: 10.5220/0011926900003612

in Bibtex Style

@conference{isaic22,
author={Sboev Alexander and Roman Rybka and Aleksandr Naumov and Anton Selivanov and Artem Gryaznov and Ivan Moloshnikov},
title={An Accuracy Comparison of the Joint and Sequential Approaches for End-to-End Related Named Entities Extraction in the Texts of Russian-Language Reviews Based on Neural Networks},
booktitle={Proceedings of the 3rd International Symposium on Automation, Information and Computing - Volume 1: ISAIC},
year={2022},
pages={348-353},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0011926900003612},
isbn={978-989-758-622-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 3rd International Symposium on Automation, Information and Computing - Volume 1: ISAIC
TI - An Accuracy Comparison of the Joint and Sequential Approaches for End-to-End Related Named Entities Extraction in the Texts of Russian-Language Reviews Based on Neural Networks
SN - 978-989-758-622-4
AU - Alexander S.
AU - Rybka R.
AU - Naumov A.
AU - Selivanov A.
AU - Gryaznov A.
AU - Moloshnikov I.
PY - 2022
SP - 348
EP - 353
DO - 10.5220/0011926900003612
PB - SciTePress