Named Entity Recognition Using BioBERT
Shubhangi C. Tirpude, Vedant R. Shahu, Meghal G. Neema, Rishika Chatterjee, Piyush H. Wanve · 2024
The focus of this study is on the crucial function of Named Entity Recognition (NER) in obtaining critical medical insights from Electronic Health Record (EHR) databases, thereby assisting clinical research and decision-making. The project focuses on leveraging the BioBert model’s capabilities to improve NER in the domain of EHR data while utilizing the unique problems and complexities offered by the N2C2 dataset. The work demonstrates BioBert’s ability to address the language intricacies and medical terminologies inherent in EHR narratives, which was accomplished through painstaking model fine-tuning and adaption to the NER problem. Comprehensive studies show that BioBert outperforms standard approaches in detecting complicated medical entity categories and adapting contextual differences found in EHR documents. The study goes beyond the predictability of BioBert’s predictions, revealing insights into its decision-making processes and contextual knowledge. In conclusion, this study demonstrates the BioBert model’s effectiveness as a robust approach for strengthening NER within the complex terrain of EHR datasets. The findings not only expand NER methodology in the healthcare industry, but also establish the framework for future research, such as domain adaption, fine-tuning procedures, and possible integration with clinical decision support systems.