Automatic Summarization of Doctor-Patient Encounter Dialogues Using Large Language Model through Prompt Tuning
Researchers have developed a novel approach to automatically summarize doctor-patient dialogues using a generative clinical large language model called GatorTronGPT. This method employs prompt-tuning techniques, which are computationally efficient as they do not require updating the LLM's parameters. Experiments on the MTS-DIALOG benchmark dataset demonstrated that the GatorTronGPT-20B model outperformed a T5-based fine-tuning solution across all evaluation metrics, highlighting the efficacy of prompt-tuned generative clinical LLMs for clinical automatic text summarization. AI
IMPACT Demonstrates efficient LLM application for clinical text summarization, potentially reducing clinician workload.