Research On Event Extraction Of Chinese Clinical Guidelines Based On Deep Learning

Clinical guidelines are the results of standardized treatment of medical knowledge,which can guide clinical diagnosis and treatment in a multidisciplinary and systematic way,and can assist medical personnel to make reasonable decisions in clinical diagnosis and disease treatment.As an important component of clinical guidelines,clinical guideline events are of great significance for disease diagnosis and treatment.This study extracts clinical guideline events and forms a series of standardized and standardized texts,which can provide information support for clinical diagnosis and drug retrieval,reduce medical omissions caused by insufficient knowledge of medical staff,and improve the overall level of clinical diagnosis and treatment.In this study,event extraction is divided into two tasks of "trigger word identification" and "argument identification".Through analyzing the text of medical guidelines,the basic event model in clinical guidelines is constructed,including model name,trigger word type,event elements in the model,etc.According to the constructed event model,the corpus is manually annotated,the text is expressed in the form of word vectors,and the word vectors are respectively input into a bidirectional long-term and short-term memory network and a convolution neural network.The learning features and higher-level features are automatically updated by using the algorithm model,iterative algorithm parameters are continuously updated in reverse,and complete two sets of algorithm models are established.Finally,the model is used to automatically extract the Chinese medical guide texts in the test set and evaluate the extraction effect.The results show that compared with the traditional machine learning model,the Bi-LSTF+CRF-based Chinese clinical guideline extraction model has obvious improvement in trigger word recognition and argument role recognition.Compared with the traditional machine learning model,CNN-based Chinese clinical guideline extraction model also has better effect on trigger word recognition and argument role recognition.