摘要
目的利用知识图谱对藏医理论进行数字化研究,在保护和推广藏医诊疗和藏文化的同时,也为进一步的应用开发打下基础。方法研究将知识图谱应用于藏医领域,并结合大模型回复和知识图谱匹配开发智能问诊系统。结果研究依托《四部医典》,以提取出的506种疾病和14种藏医特色疗法为中心构造知识图谱,对疾病诊断研究与配伍进行可视化展示。构造文本量达20万字的藏医命名实体识别数据集和藏医问诊意图识别数据集。采用rbt3-BiLSTM-crf和bert+textcnn方法进行训练,模型效果分别达到74.26%的f1 score和0.98714的准确率。结论研究构建的知识图谱展现了《四部医典》对疾病的讨论情况。侧面反映了高原居民在生活居住和饮食方面可能引发的疾病问题。通过人工智能对藏医文本进行电子化,构建藏医领域的问诊系统。
Objective As an important part of Chinese ethnic medicine,Tibetan medicine integrates a variety of medical theories,covering rich drug resources and unique drug preparation techniques.Using the knowledge graph to conduct digital research on Tibetan medicine theory not only protects and promotes Tibetan medicine diagnosis and Tibetan culture,but also lays the foundation for further application and development.Methods The research applied the knowledge graph to the field of Tibetan medicine,and developed an intelligent diagnosis system by combining large model responses and knowledge graph matching.Results The study relied on the Four Medical Classics to construct a knowledge map centered on 506 diseases and 14 Tibetan medicine-specific therapies,providing a visual display of disease diagnosis research and compatibility.Construct a Tibetan medicine named entity recognition dataset and a Tibetan medicine diagnosis intent recognition dataset with a text volume of 200000 words.The rbt3-BiLSTM-crf and bert+textcnn methods were used for training,and the model achieved an f1 score of 74.26%and an accuracy of 0.98714,respectively.Conclusion The knowledge map constructed in this study shows the discussion of diseases in the Four Medical Classics.It reflects the potential disease problems that may arise in the living and diet of plateau residents.Through artificial intelligence,the Tibetan medical texts are digitized to build a diagnosis system in the field of Tibetan medicine.
作者
周泽先
刘晓静
ZHOU Zexian;LIU Xiaojing(Department of Computer Technology and Application,Qinghai University,Xining 810016,China;Qinghai Provincial Key Laboratory of Media Convergence Technology and Communication,Xining 810016,China)
出处
《现代中医药》
2025年第4期27-34,共8页
Modern Chinese Medicine
基金
国家自然科学基金(61862053)
青海省科技厅应用基础研究项目(2021-ZJ-717)。
关键词
知识图谱
藏医学
可视化分析
智能问答系统
自然语言处理
knowledge graph
tibetan medicine
visualization analysis
intelligent question-answering system
natural language processing
作者简介
通讯作者:刘晓静,硕士,教授。E-mail:645020710@qq.com。
