Advancing Clinical Text Summarization through Extractive Methods using BERT-Based Models on the NBME Dataset
R Sudarshan, D. Sasikala, S. Kalavathi · 2023
Medical notes contains a vast amount of healthcare information. NLP techniques are essential for extracting the most relevant insights from abundance data, enabling healthcare professionals to make well-informed decisions.In the study, clinical text feature extraction was performed on the NBME dataset for extractive summarization using PubMedBERT, BioBERT, and DeBERTa. It was found that PubMedBERT achieved 92% accuracy, BioBERT achieved 95% accuracy, and DeBERTa achieved 86% accuracy. PubMedBERT and BioBERT performed well on clinical text summarization task these models have a slight edge over DeBERTa. These results suggest that domain-specific pretrained language models like PubMedBERT and BioBERT are useful for extractive summarization of clinical text. Performing feature extraction for extractive summarization of medical notes can make healthcare information more concise and accessible.Facilitates physicians and other healthcare professionals to improve their analysis on medical reports.Overall,advances in extractive summarisation of clinical text through NLP can lead to better-informed healthcare professionals and improved patient outcomes.