浏览全部资源
扫码关注微信
1.上海中医药大学附属曙光医院内分泌科(上海 201203)
2.上海中医药大学信息化办公室(上海 201203)
3.东南大学计算机科学与工程学院(江苏 南京 211189)
4.斯威本科技大学科学、计算和工程技术学院(澳大利亚 墨尔本 3122)
5.江苏省人民医院(南京医科大学第一附属医院)信息处(江苏 南京 210029)
6.东南大学新一代人工智能技术与交叉应用教育部重点实验室(江苏 南京 211189)
龚凡,女,博士,副主任医师,主要从事中医药防治内分泌代谢疾病临床及交叉研究工作;
景慎旗,正高级工程师,硕士研究生导师;E-mail:jingshenqi@jsph.org.cn
吴天星,副教授#,博士研究生导师; E-mail:tianxingwu@seu.edu.cn
收稿日期:2024-10-09,
纸质出版日期:2025-06-10
移动端阅览
龚凡,刘波,沙航宇,等.基于大语言模型检索增强的中医“药食同源”饮食建议生成方法[J].上海中医药杂志,2025,59(6):1-8.
GONG Fan,LIU Bo,SHA Hangyu,et al.A method for generating dietary recommendations following "food and medicine homology" principle of traditional Chinese medicine using retrieval‑augmented large language models[J].Shanghai Journal of Traditional Chinese Medicine,2025,59(6):1-8.
龚凡,刘波,沙航宇,等.基于大语言模型检索增强的中医“药食同源”饮食建议生成方法[J].上海中医药杂志,2025,59(6):1-8. DOI: 10.16305/j.1007-1334.2025.z20241009011.
GONG Fan,LIU Bo,SHA Hangyu,et al.A method for generating dietary recommendations following "food and medicine homology" principle of traditional Chinese medicine using retrieval‑augmented large language models[J].Shanghai Journal of Traditional Chinese Medicine,2025,59(6):1-8. DOI: 10.16305/j.1007-1334.2025.z20241009011.
目的
2
研究基于大语言模型检索增强技术生成符合中医“药食同源”原则的饮食建议的方法,提升饮食建议生成效率和质量,以满足不同健康状况用户的个性化饮食需求。
方法
2
首先设计一套统一的提示工程指令,对ChatGPT 4.0、通义千问、文心一言等主流大语言模型进行对比测试。基于测试结果选择表现最佳的大语言模型作为基座模型。接着进一步引入检索增强生成技术,通过稠密检索从知识库中获取与问题相关的文档片段,以增强大语言模型生成的饮食建议的专业性和准确性。随后比较检索增强生成的效果,并进行评价。
结果
2
研究结果显示,通义千问在评估中表现最佳。引入检索增强生成技术后,模型生成的答案相关性显著提升,从0.81提高到0.93。实际案例分析也表明,检索增强生成技术能够有效提高大语言模型生成的饮食建议的中医理论合理性,确保饮食建议既科学又满足个性化要求。
结论
2
基于大语言模型检索增强技术生成的个性化饮食建议符合中医“药食同源”理念和用户的个性化需求。检索增强生成技术提升了模型生成内容的相关性、忠实度和上下文相关性,确保了饮食建议的专业性和实用性。所提出的方法不仅提高了饮食建议的质量,还为个性化医疗的实践提供了新的工具和视角。未来研究需要进一步优化模型的中医知识理解能力,提高中医知识库的质量,以提供更加全面和准确的健康饮食建议。
Objective
2
To develop a method for generating personalized dietary recommendations that align with the "food and medicine homology" principle of traditional Chinese medicine (TCM) using retrieval-augmented large language models (LLMs), and to improve the efficiency and quality of dietary recommendations for individuals with varying health conditions.
Methods
2
Initially, a unified set of prompt engineering instructions was designed to compare the performance of leading LLMs such as ChatGPT 4.0, Qwen, and ERNIE Bot. Based on the test results, the best-performing LLM was selected as the base model. Subsequently, a retrieval-augmented generation (RAG) technique was incorporated to enhance the professionalism and accuracy of dietary recommendations generated by the LLM. This involved dense retrieval from a knowledge base to obtain relevant document segments related to the query. The effectiveness of RAG was then compared and evaluated.
Results
2
Experimental results indicated that Qwen outperformed other models in the evaluation. After introducing the RAG, the relevance of the model-generated answers significantly improved, with the score rising from 0.81 to 0.93. Case analyses further demonstrated that RAG effectively improved the theoretical rationality of TCM-based dietary recommendations generated by the LLMs, ensuring both scientific rigor and personalization.
Conclusions
2
Personalized dietary recommendations generated using retrieval-augmented LLMs conform to the "food and medicine homology" principle of TCM and cater to individual needs. RAG enhances the relevance, fidelity, and contextual consistency of the model's output, ensuring the professionalism and practicality of dietary recommendations. The proposed method not only improves the quality of dietary recommendations but also provides new tools and perspectives for the practice of personalized healthcare practices. Future research should focus on optimizing the model's understanding of TCM knowledge and improving the quality of the TCM knowledge base to offer more comprehensive and precise dietary recommendations.
单峰 , 黄璐琦 , 郭娟 , 等 . 药食同源的历史和发展概况 [J]. 生命科学 , 2015 , 27 ( 8 ): 1061 - 1069 .
杨上善 . 黄帝内经太素 [M]. 王洪图,点校. 北京 : 科学技术文献出版社 , 2005 : 17 .
杨叔禹 . 国家糖尿病基层中医防治管理指南(2022) [J]. 中医杂志 , 2022 , 63 ( 24 ): 2397 - 2414 .
唐雪阳 , 谢果珍 , 周融融 , 等 . 药食同源的发展与应用概况 [J]. 中国现代中药 , 2020 , 22 ( 9 ): 1428 - 1433 .
OpenAI . Introducing ChatGPT [EB/OL].[ 2022-11-30 ]. https://openai.com/index/chatgpt https://openai.com/index/chatgpt .
THIRUNAVUKARASU A J , TING D S J , ELANGOVAN K , et al . Large language models in medicine [J]. Nat Med , 2023 , 29 ( 8 ): 1930 - 1940 .
FAN W , DING Y , NING L , et al . A survey on RAG meeting LLMs: Towards retrieval-augmented large language models [C]//ACM. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining . Barcelona : ACM , 2024 : 6491 - 6501 .
佚名 . 黄帝内经·素问 [M]. 广州 : 广东科技出版社有限公司 , 2022 : 65 .
孙晓生 . 《食疗本草》的学术成就及现代应用 [J]. 新中医 , 2011 , 43 ( 5 ): 129 - 131 .
叶橘泉 . 食物中药与便方 [M]. 南京 : 江苏科学技术出版社 , 1980 .
钱伯文 . 中国食疗学 [M]. 上海 : 上海科学技术出版社 , 1987 .
彭铭泉 . 中国药膳学 [M]. 北京 : 人民卫生出版社 , 1985 .
阙灵 , 杨光 , 李颖 , 等 . 《既是食品又是药品的物品名单》修订概况 [J]. 中国药学杂志 , 2017 , 52 ( 7 ): 521 - 524 .
国家中医药管理局综合司 , 国家药品监督管理局综合司 . 国家中医药管理局综合司 国家药品监督管理局综合司关于发布《古代经典名方关键信息表(“竹叶石膏汤”等25 首方剂)》的通知 [EB/OL].( 2023-07-28 )[ 2024-09-10 ]. http://www.natcm.gov.cn/kejisi/gongzuodongtai/2023-07-28/31404.html http://www.natcm.gov.cn/kejisi/gongzuodongtai/2023-07-28/31404.html .
MARVIN G , HELLEN N , JINGO D , et al . Prompt engineering in large language models [C]//Springer Nature Singapore. Proceedings of international conference on data intelligence and cognitive informatics . Singapore : Springer Nature Singapore , 2023 : 387 - 402 .
Google . Gemini [EB/OL]. [ 2024-09-25 ]. https://gemini.google.com https://gemini.google.com .
阿里云 . 通义千问 [EB/OL]. [ 2024-09-25 ]. https://tongyi.aliyun.com https://tongyi.aliyun.com .
百度 . 文心一言 [EB/OL].[ 2024-10-01 ]. https://yiyan.baidu.com https://yiyan.baidu.com .
ZHANG H , CHEN J , JIANG F , et al . HuatuoGPT, towards taming language model to be a doctor [C]//ACL SIGDAT. Proceedings of Findings of the Association for Computational Linguistics: EMNLP 2023 . Singapore : ACL SIGDAT , 2023 : 10859 - 10885 .
科大讯飞 . 讯飞晓医 [EB/OL]. [ 2024-11-08 ]. https://xinghuo.xfyun.cn/desk https://xinghuo.xfyun.cn/desk .
DEVLIN J , CHANG M , LEE K , et al . BERT: Pre-training of deep bidirectional transformers for language understanding [C]// NAACL . Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics : Human Language Technologies. Minneapolis: NAACL , 2019 : 4171 - 4186 .
杨飞洪 . 面向中文临床自然语言处理的BERT模型研究 [D]. 北京 : 北京协和医学院 , 2021 .
KARPUKHIN V , OGUZ B , MIN S , et al . Dense passage retrieval for open-domain question answering [C]//ACL SIGDAT. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) . Hong Kong : ACL SIGDAT , 2020 : 6769 - 6781 .
ES S , JAMES J , ANKE L E , et al . RAGAs: Automated evaluation of retrieval augmented generation [C]// EACL . Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics : System Demonstrations.Bergen: EACL , 2024 : 150 - 158 .
0
浏览量
0
下载量
0
CSCD
0
CNKI被引量
关联资源
相关文章
相关作者
相关机构