Research Of Medical Named Entity Recognition And Development Of Electronic Medical Record Marking System

Few years recently,artificial intelligence has been developed rapidly and integrated into various fields.As an important part of our society,the medical field is trying to combine with artificial intelligence these years.A typical application of this integration is Medical Named Entity Recognition(MNER).There are many critical problems to be addressed in MNER.First of all,public data sets wildly used by researchers are too simple to make a good simulation of real medical scenarios.Moreover,research on MNER lags behind research on general field.There is still some experience that can be transformed from the general field.This paper did some work to try to solve the above questions.Firstly,the paper introduces the main public data sets and points out their problems.Secondly,a new data set is developed based on electric medical records.Thirdly,two different baseline models,BiLSTM-CRF and Transformer Encoder-CRF,are tested in that data set.Then,based on the problems of the baseline models,three improvements are proposed.Finally,after performance improvement,an electronic medical record auxiliary marking system is designed and developed to resolve the problem in the labeling step.In the chapter of model improvement,the accuracy of the model is improved from 90.78%to 92.68%by the join action of three methods.In the chapter of system design and development,this paper designs and implements a functional and easy-to-use electronic medical record auxiliary marking system,and tests the functions of the system.