从传统自然语言处理到大语言模型电子病历生成技术的探索  

Exploration and Research of Electronic Medical Record Generation Technology from Traditional Natural Language Processing to Large Language Model

作　　者：牛宇翔 葛珊衫 王力华 NIU Yuxiang;GE Shanshan;WANG Lihua(Capital Medical University,Beijing Friendship Hospital,Beijing 100050,China)

机构地区：[1]首都医科大学附属北京友谊医院,北京100050

出　　处：《医学信息学杂志》2025年第3期79-83,101,共6页Journal of Medical Informatics

摘　　要：目的/意义探讨并展望大语言模型应用于病历文书生成的效果和关键技术,以提升临床工作效率。方法/过程综述病历文书生成技术发展历程,从传统自然语言处理方法到深度学习方法,再到大语言模型创新应用,并探讨关键技术路线。结果/结论未来研究方向主要包括基于上下文学习的病历文书生成、基于检索增强生成的病历文书生成及基于混合专家模型的病历文书生成技术。Purpose/Significance To discuss and look forward to the effect and key technologies of applying large language model to medical record generation,so as to improve the efficiency of clinical work.Method/Process The paper reviews the development of medical record generation technology,from traditional natural language processing(NLP)methods to deep learning methods,and then to innovative applications of large language models(LLM),and discusses the key technical routes.Result/Conclusion The future research direction mainly includes the generation of medical records based on context learning,the generation of medical records based on retrieval-augmented generation and the generation of medical records based on mixture of experts.

关 键 词：大语言模型 病历生成 自然语言处理 医疗信息化 

分 类 号：R-058[医药卫生]