基于命名实体识别的医学病历自动生成研究与实现

AbstractIn recent years,the rapid development of "Internet medical"technology has made the managementof hospitals more automated and networked,and the traditional manual paper medical records are beingreplaced by electronic medical records.Electronic medical records can transmit information more quickly,making the diagnosis and treatment process more convenient and easy to manage.Through years ofdevelopment,the usage of electronic medical records have been gradually expanded in practicalapplications.At present,although the major hospitals have already put into the use of electronic medicalrecord system,in most cases it still need the doctors to operate step by step,and sometimes even usemanual medical record entry system,there is no really effective reduction of the workload of doctors.Especially for senior doctors,cumbersome operation will also affect work efficiency.Therefore,in order tosimplify the work of doctors'medical records,this paper studies the automatic generation system ofelectronic medical records.In this paper,the author studies on electronic medical records automatically generatedby named entityrecognition technology,propose the named entity recognition method based on neural network (BILSTM-CRF)on the extraction of symptoms and disease model of important medical data such as training.Bycomparing the effect of word and character level BILSTM-CRF model,more suitable for named entityrecognition of medical text input methods are choosed;The method of short text classification is used topredict the unknown symptoms and predict their disease category.By comparing the classification effect oftextCNN classification algorithm and textRNN classification algorithm,this paper explores the advantagesand disadvantages of the two classification algorithms in the application process,and selects the betterclassification algorithm to be implemented in the system.Then it introduces the formulation of medical record template and summarizes the specificinformation contained in outpatient medical record.By analyzing a large number of outpatient medicalrecords information,the structure between different levels is summarized,and an electronic medical recordtemplate is created.Finally,the method ofrule matching is used to automatically generate medical records,and the medical records automatic generation system is realized.Keywords:electronic medical record,named entity recognition,neural network,text classification