基于自动文本分类的中医症状标准化路径探讨  

Discussion on the path of TCM symptom standardization based on automatic text classification

作　　者：赵宗耀 陈剑北 柳辰玥 薛哲 刘骐瑞 陈家旭 ZHAO Zong-yao;CHEN Jian-bei;LIU Chen-yue;XUE Zhe;LIU Qi-rui;CHEN Jia-xu(School of Traditional Chinese Medicine,Beiing University of Chinese Medicine,Beiing 100029,China)

机构地区：[1]北京中医药大学中医学院,北京100029

出　　处：《中华中医药杂志》2022年第10期5671-5675,共5页China Journal of Traditional Chinese Medicine and Pharmacy

摘　　要：基于“六异识候”症状标准化理论体系,明确了中医症状标准化的3个阶段任务和基于人工智能的技术路径。第1阶段任务是明确症状类型;第2阶段任务是明确不同症状类型描述公式中的具体内容;第3阶段任务是构建标准化症状数据库以及各类标准化症状描述公式数据库。这3个阶段任务均可以采用人工智能技术自动完成,技术路径是:文本分类-文本理解-知识提取。本课题组提出了一个基于LFText与TextCNN的自动文本分类模型来自动化症状分类(第1阶段任务),模型整体效果评价指标:分类别平均准确率(MAP)、宏平均F1值(Macro F1)、宏平均准确率(Macro Accuracy)分别是0.8107、0.8117、0.8290,证明了本课题组提出的中医症状标准化路径的有效性,并为后续两阶段的标准化任务奠定了数据与技术基础。Based on the symptom standardization theoretical system of the ‘six different syndromes’, the three-stage task of TCM symptom standardization and the technical path based on artificial intelligence technology are clarified. The first stage task is to clarify the types of symptoms;the second stage task is to clarify the specific content of the description formulas for different symptom types;the third stage task is to build a standardized symptom database and various standardized symptom description formula databases. These three-stage tasks can all be completed automatically using artificial intelligence technology.The technical path is text classification-text understanding-knowledge extraction. We propose an automatic text classification model based on LFText and TextCNN to automate symptom classification. MAP, Macro F1, and Macro Accuracy are 0.8107,0.8117, and 0.8290 respectively, which proves the effectiveness of the standardization path of TCM symptoms we proposed.It also lays a data and technical foundation for the standardization tasks in the subsequent two phases.

关 键 词：六异识候 症状 标准化 文本分类 

分 类 号：R241[医药卫生—中医诊断学]