Research And Application Of Electronic Medical Record Information Extraction Based On Machine Learning

With the development of artificial intelligence and machine learning,medical information systems are becoming more and more perfect.Electronic Medical Record(EMR),as an important information flow carrier,has also become a very important information resource in the information system.At present,the information extraction of electronic medical records is an important research direction of text data mining,which is of great significance for auxiliary diagnosis and medical information query in the medical field.Therefore,this paper takes the structured text as the starting point,and first introduces the research status and development of named entity recognition and relationship extraction.It also details the principles of artificial neural networks,language models,and transformer frameworks.Secondly,in the work of electronic medical naming entity recognition,the medical text is added to the language model BERT for pre-training finetuning,and the self-attention mechanism is introduced,and the two-way LSTM model is integrated for feature extraction,and the named entity recognition is carried out by random condition field as a classification constraint.In the labeling data of CCKS in 2019,the model proposed in this paper achieved a recall rate of 79.79%,the precision rate of 79.49%and an F1 score of 79.64% compared with the traditional named entity recognition model.Experimental results show that the model can effectively carry out the task of identifying medical named entities.Then,in the work of electron medical record entity relationship classification,this paper proposes a model method that fuses convolutional operations and LSTM modules to form Conv Bi LSTM for text feature extraction.Combined with the pre-trained language model,the feature fusion is performed by using the maximum pooling operation to complete the relationship classification task.In this paper,the precision of the model on the entity relationship dataset of MMC Chinese diabetes research literature of Ruijin Hospital was 79.19%,the recall rate was 81.22%,and the F1 score was 80.19%.