摘要
中医知识的现代化与现代西医学知识可以建立深度的互通互解。在大语言模型背景下,以传统中医和现代西医的概念关系理解为切入点,该文提出一种基于精调LLaMA模型的中西医概念关系对比分析方法。研究中,首先选定中西医中一组相通的基本概念术语,并进行了相应的文本数据集构建;随后基于LLaMA模型分别对两个数据集进行精调学习,得到关于同组基本概念术语的两个大语言模型;其次,基于基本概念术语集和两个文本数据集,设计了一套有关概念术语知识的填空和问答题集自动生成方法,并由训练得到的两个模型分别作答;最后,依据两个模型的作答结果,采用自动化比对和人工辅助判别的方法,进行概念术语的一致性或差异性理解分析。实验结果表明,精调LLaMA模型能够对构造的文本数据集进行有效的建模理解;而作答结果对比分析显示,在基本中西医术语概念关系的理解上,两个模型约70%呈现一致性,但也有近30%的测试理解存在不同。从中可知,传统中医知识在现代化过程中,与现代西医知识已有较深度的融合,但其中仍有较多的基本概念术语未能与现代医学知识建立有效的连接互通。
It is argued that the modernization of traditional Chinese medicine required the establishment of a deep and mutual interchange with Western medical knowledge.This paper proposes a contrastive study on concept relationships between traditional Chinese medicine and Western medicinevia a finetuned LLaMA.In the research,a set of common basic concept terms in both traditional Chinese and Western medicine is selected,and corresponding text datasets are collected.The LLaMA model is fine-tuned by the two datasets,respectively,resulting two large language models for the basic concept terms within each group.Further,a method for generating a set of fill-in-the-blank and question-answer test questions about concept term knowledge is designed.Finally,automated comparison and manual assessment are employed to analyze the consistency or differences in the understanding of concept terms provided by the two finetuned LLaMa models.Experimental results demonstrated that,in the understanding of the relationships between basic concepts in both traditional Chinese and Western medicine,the two models exhibited approximately 70%consistency.The results suggest that while traditional Chinese medicine knowledge had experienced a deep integration with Western medical knowledge in the process of modernization,there are still many basic concept terms that had not established effective connections for mutual exchange with Western medical knowledge.
作者
叶淋潮
邵会会
谢振平
YE Linchao;SHAO Huihui;XIE Zhenping(School of Artificial Intelligence and Computer Science,Jiangnan University,Wuxi,Jiangsu 214122,China;Jiangsu Key Laboratory of Media Design and Software Technology,Wuxi,Jiangsu 214122,China)
出处
《中文信息学报》
北大核心
2025年第2期162-170,共9页
Journal of Chinese Information Processing
基金
国家自然科学基金(62272201)。
关键词
中西医差异分析
术语概念关系
问题自动生成
LLaMA模型
differential analysis of traditional Chinese and western medicine
terminology concept relationships
automated question generation
LLaMA Model
作者简介
叶淋潮(1999-),硕士研究生,主要研究领域为自然语言处理。E-mail:6213113131@stu.jiangnan.edu.cn;邵会会(1995-),博士研究生,主要研究领域为知识计算与机器学习。E-mail:7223115004@stu.jiangnan.edu.cn;通信作者:谢振平(1979-),博士,教授,主要研究领域为认知计算。E-mail:xiezp@jiangnan.edu.cn。
