Enhancing Clinical Note Generation from Doctor-Patient Conversations through Semantic Partition-Oriented Summarization

Clinical note generation from doctor-patient con-versations is an essential task that helps to maintain the medical records of patients. The process of writing clinical notes is a time-consuming task for doctors, and redundant or inaccurate information in clinical notes may have adverse consequences. In this paper, we propose a novel approach to clinical note generation. Our contribution lies in proposing a semantic partitioning and clustering method for the extractive module of this task. We show that our semantic-based partition method provided a way to extract relevant information for specific sections. This enables large language models to produce high-quality clinical notes and give accurate information for each section. Our approach outperformed other systems in ROUGE-1 score on the MEDIQA-Sum 2023 Shared Task dataset.