KEMedGPT: Intelligent Medical pre-consultation with Knowledge-Enhanced Large Language Model

Abstract

Large language models(LLMs) are driving productivity advancements in the fields of medical healthcare and information systems. Existing medical LLMs types vary widely and frequently pose challenges for small and medium-sized enterprises(SMEs) to deploy. To address these limitations, we propose KEMedGPT: a knowledge-enhanced medical GPT specifically designed to improve the medical knowledge and consultation capabilities of LLMs. KEMedGPT employs a two-stage strategy and leverages the unique medical text Q&A data from an Internet medical society for training. This approach simulates human-like decision-making processes using real-world patient data, enhancing the model's relevance and applicability. Our experiments demonstrate that KEMedGPT excels in multi-turn dialogue for pre-consultation, effectively facilitating interactive exchanges that enhance the early identification of patient needs and the delivery of personalized medical advice. This capability significantly improves medication safety and elevates the overall quality of healthcare services. Extensive and rigorous evaluations of the model highlight KEMedGPT's superiority, outperforming existing general and specialized large language models.

Ähnliche Arbeiten