Research On Clinical Symptom Data Elements Of Traditional Chinese Medicine Based On Machine Learning

From ancient times to the present,the medical records,classic works and clinical records of traditional Chinese medicine are mostly described by natural language.This leads to the nonstandard description of Chinese medicine terms.Symptoms of TCM,as the main part of the clinical information,are non standard such as synonym,complete homonyms,non-criterion definition,Small scale of quantitative region,and so on.From the perspective of data sharing and exchange,this paper uses machine learning of the Natural Language Processing technology,to standardize the symptom terms from the name,definition,data type and range of symptoms by extracting the clinical symptom data element of traditional Chinese medicine.Metadata is a universal data standard used for sharing,exchanging and managing at present.Data element is the smallest data unit in metadata management,and it is an abstract description of the concepts and relationsh ips in the real world.The standardization of data elements includes data definition,type,representation and synonym merging.The clinical symptom of traditional Chinese medicine is the core concept of diagnosis information of TCM,and it is also the bas ic data of the knowledge base of TCM in the health information management system.According to the results of this study,it is a feasible and efficient method to standardize symptom data from the perspective of knowledge engineering by standardizing the s ymptom terms based on the data element standard,by using machine learning.According to the characteristics of clinical records of Chinese medicine,using the Natural Language Processing technology based machine learning to extract from the text,we can dynamically extract and update the data elements of clinical symptoms of traditional Chinese medicine.It effectively maintains the characteristics of personalized medical records written by doctors in natural language.The main research work and innovation in this paper are as follows:(1)This paper have studied the characteristics of TCM clinical symptom terms,the relationship between symptoms and symptoms,the properties of symptoms,and made a superficial exploration of the relationship between sympto ms and syndromes.(2)This paper has studied the data element definition,standard model,and mode of management.According to TCM clinical information system application status and the clinical characteristics of TCM symptoms,extraction methods and princ iples of Chinese medicine clinical symptoms data elements have been analyzed.And from these,a three-layer extraction model corresponding to the p urposes has been proposed.(3)This paper has studied the term symptom recognition technology.By analyzing the characteristics of more than 1900 clinical records,the clinical symptoms terms have been extracted from clinical records texts with HMM and CRF technologies.In the process of extracting the symptom terms,most information of symptoms have been saved,which is benefit to subsequent extraction of symptom data elements.(4)This paper have studied the high efficient data structure of symptom terminology,and constructed multilevel glossary structure,which can satisfy the needs of term recognition,data element extraction and data mining.(5)This paper have studied technology roadmap of extraction of symptom data element from clinical medical records text and literature with machine learning technology,and completed part of symptom data elements extracti on.