Medical Named Entity Recognition Research Based On Deep Learning

With the construction of electronic information in the medical field,the amount of text data such as medical records has increased exponentially.The medical industry could collect the medical information of patients more quickly and conveniently with the development of Internet information system,,but the medical electronic medical record data still has the characteristics of heterogeneity,distribution and fragmentation.In order to mine valuable analysis and information mining from massive medical data,medical data should be structured.Through the application of Natural Language Processing,the named entity recognition of text data in the medical field will lay a foundation for the structured representation of data.And the high quality named entity data is the indispensable basic link to build the intelligent medical system.However,high-quality physical data is usually manually annotated by experts,which is costly and time-consuming.Therefore,it is an indispensable link in the field of medical text data mining to use an effective computer algorithm to identify named entities and improve the accuracy of the algorithm.Medical Named Entity Recognition is different from the traditional Named Entity Recognition,which pays more attention to the entities such as symptoms,organs,treatment methods,drugs and diseases.However,due to the lack of standard nomenclature methods in the medical field,medical nomenclature entity recognition can rarely achieve very satisfactory results with models,so it is still a very difficult task to identify medical nomenclature entity.This paper firstly introduces the research of Medical Named Entity Recognition recent year,and mainly summarizes the development process of Named Entity Recognition based on deep learning model.Based on the standard deep learning model LSTM-CRF,this paper puts forward SE-LSTM which modifies gating mechanism of LSTM,and improves the accuracy of recognition effectively.Besides,This paper applies the latest research progress in the field of natural language processing to the field of medical name entity recognition(BERT and XLNet).The experimental results show that the F1 value improved by 1-3%through modifying gating mechanism of LSTM.When BERT and XLNet are used,there is a significant improvement in larger data set.