Design And Implementation Of Medical Records Writing Assistant System Based On Named Entity Recognition

Electronic medical record is an important tool for medical staff to carry out medical activities,and it is of great significance in assisting doctors to work.Inpatient medical record is an important part of electronic medical records,including all kinds of medical information generated in the process of hospitalization.To ensure the quality and input efficiency of inpatient medical records,is very important.However,the current widely used electronic medical records lack the corresponding auxiliary function in terms of free text entry in inpatient medical records.In order to improve the interactivity and operability of medical records input,reduce the complexity of writing,reduce doctor load and improve work efficiency,this paper uses the conditional random field algorithm to identify the named entity of electronic medical records.It helps to optimize and improve the entry mode in medical records.First of all,on the basis of conditional random field algorithm and some comparative tests of different feature templates,this paper obtains the best recognition model,completes model training for corpus and performance evaluation,and help to extract important named entity in medical records,where a good recognition effect could be achieved.Then,this paper extracts feature words through named entity recognition,analyses the latent semantic using the latent Dirichlet distribution model,compares the similarity between the query and the medical records in the semantic level,and recommends the high similarity of medical records as input reference.In addition,for the medical records which need the existing clinical data of patients,this paper intuitively presents patient's medical data during the hospitalization in the form of timeline,where relatively structured data could be quickly inserted according to user defined format.Medical documents with many free text need to be processed through the entity recognition or a certain degree of information extraction so that users could insert the required information.