OAK

Re-Ranking System with BERT for Biomedical Concept Normalization

Metadata Downloads
Abstract
In recent years, various neural network architectures have been successfully applied to natural language processing (NLP) tasks such as named entity normalization. Named entity normalization is a fundamental task for extracting information in free text, which aims to map entity mentions in a text to gold standard entities in a given domain-specific ontology; however, the normalization task in the biomedical domain is still challenging because of multiple synonyms, various acronyms, and numerous lexical variations. In this study, we regard the task of biomedical entity normalization as a ranking problem and propose an approach to rank normalized concepts. We additionally employ two factors that can notably affect the performance of normalization, such as task-specific pre-training (Task-PT) and calibration approach. Among five different biomedical benchmark corpora, our experimental results show that our proposed model achieved significant improvements over the previous methods and advanced the state-of-the-art performance for biomedical entity normalization, with up to 0.5% increase in accuracy and 1.2% increase in F-score.
Author(s)
Cho, HyejinChoi, DonghaLee, Hyunju
Issued Date
2021-08
Type
Article
DOI
10.1109/ACCESS.2021.3108445
URI
https://scholar.gist.ac.kr/handle/local/11346
Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
Citation
IEEE ACCESS, v.9, pp.121253 - 121262
ISSN
2169-3536
Appears in Collections:
Department of AI Convergence > 1. Journal Articles
공개 및 라이선스
  • 공개 구분공개
파일 목록

Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.