Re-Ranking System with BERT for Biomedical Concept Normalization

In recent years, various neural network architectures have been successfully applied to natural language processing (NLP) tasks such as named entity normalization. Named normalization is a fundamental task for extracting information in free text, which aims map mentions text gold standard entities given domain-specific ontology; however, the biomedical domain still challenging because of multiple synonyms, acronyms, and numerous lexical variations. this study, we regard ranking problem, propose an approach rank normalized concepts. We additionally employ two factors that can notably affect performance normalization, task-specific pre-training (Task-PT) calibration approach. Among five different benchmark corpora, our experimental results show proposed model achieved significant improvements over previous methods advanced state-of-the-art with up 0.5% increase accuracy 1.2% F-score.