摘要
从自然语言处理技术(NLP)入手,对比TFIDF与Word2vec方法抽取结果,探讨更适用于中医医案症状信息自动化抽取的研究思路,为发展中医医案的自动化分析提供参考。在构建好的医案词典基础上,利用TFIDF与Word2vec方法分别对心系医案数据进行症状抽取,并对结果进行对比分析。在医案中,病人发病往往伴有心悸、胸闷、胸痛、气喘、头晕等症状,且发现症状与症状之间也存在某些联系。实验评估结果表明,Word2vec方法抽取的精确率和召回率均高于TFIDF方法抽取的结果。相比TFIDF方法,将Word2vec方法应用于中医医案症状信息自动化抽取任务中效果更佳。
Starting from the natural language processing technology,this paper compares the extraction results of TFIDF and word2vec methods,and discusses the research ideas that are more suitable for the automatic extraction of symptom information of traditional Chinese medicine medical records,so as to provide reference for the development of automatic analysis of traditional Chinese medicine medical records.On the basis of the constructed medical record dictionary,TFIDF and word2vec methods are used to extract symptoms from the heart related medical record data,and the results are compared and analyzed.In medical records,patients often have palpitations,chest tightness,chest pain,asthma,dizziness and other symptoms,and there are some links between symptoms.The experimental results show that the precision and recall rate of Word2vec are higher than TFIDF.Compared with the TFIDF method,the Word2vec method is more effective in the automatic extraction of TCM medical case symptom information.
作者
屈丹丹
杨涛
胡孔法
QU Dan-dan;YANG Tao;HU Kong-fa(School of Artificial Intelligence and Information Technology,Nanjing University of Chinese Medicine,Nanjing 210023,China)
出处
《软件导刊》
2021年第2期44-48,共5页
Software Guide
基金
国家自然科学基金项目(82074580,81674099)
国家重点研发计划项目(2017YFC1703500)。