Named Entity Recognition Of Chinese Medical Records Based On Deep Learning

With the continuous development of national informatization,the combination of Internet technology and medical industry is getting closer and closer.Influenced by computer technology,the medical industry has completed the transformation from artificial to information management,and has accumulated a large number of electronic medical records.Named Entity Recognition(NER)is a basic task of natural language processing,which is used to recognize and classify related entities in unstructured text.Named Entity Recognition(NER)refers to extracting several entities and boundaries of medical concepts from electronic medical records.It can lay a foundation for the follow-up realization of medical decision-making and the construction of medical knowledge map.Due to the specialty particularity of the medical field,the medical record involves a lot of patient privacy,and the personal habits of the recorder lead to the varying degree of structuralization of the electronic medical record,which brings great challenges for researchers to explore the value and promote the development of the wisdom of the medical industry.To solve this problem,combined with the characteristics and advantages of natural language processing tasks,this paper proposes a named entity recognition method based on deep learning,which extracts medical entities from medical records,and verifies the feasibility and effectiveness of the proposed model in named entity recognition by combining public data sets.The work content of this paper mainly includes the following three parts:(1)In this paper,900 original data from the diabetes dataset of Ruijin Hospital of Tianchi Competition and CCKS(National Knowledge Atlas and Semantic Computing Conference)were pre-processed for annotation,and the differences in entity description in CCKS2017 and2019 were normalized to achieve the unification of the annotated entities of the two data sets.The annotation scheme was realized with Python language,and the medical entities involved in the experiment were annotated with {B,I,O} ternary annotations.(2)This paper uses a Bert-BILSTM-CRF Chinese medical record named entity recognition method based on pre-training model.Considering the lack of a large number of professional annotated corpora for named entity recognition tasks in the medical field,a pre-training BERT model was introduced to process word vectors,which greatly accelerated the convergence speed of downstream tasks.At the same time,BERT expressing ability for the large-scale semantic model can effectively solve the traditional language word vector expression is too single,for subsequent semantic context information sequence of short and long time memory model,the semantic sequence prediction of conditional random field model lays a foundation to the promotion of medical record entity recognition.The results show that the F1 value of the named entity recognition model based on the pre-training mechanism is 4.28% higher than that of the traditional language model,which fully indicates that the introduction of the pre-training mechanism can effectively improve the model recognition efficiency.(3)Based on the pre-trained BERT-BILSTM-CRF model,this paper proposes a deep learning model based on local feature fusion BERT-BILSTM-IDCNN-CRF by combining the iterative expansive convolutional neural network to improve the model.The results show that the F1 value of the improved entity recognition model is improved by 1.28% on the basis of the original,indicating that the expanded convolutional neural network can improve the effect of Chinese named entity recognition.(4)This paper uses Python language to design a Chinese electronic medical record online recognition system,which takes the medical record recognition model proposed in this paper as the computing core and can effectively extract entities in medical records.