Research On Entity Recognition And Normalization For Online Medical Consultation Text

The online consultation module in the medical website is one of the main carriers for providing medical knowledge,which is of great significance for clinical practice and reducing medical costs.Traditional entity recognition methods that use background knowledge cannot make full use of knowledge,thereby affecting the performance of entity recognition tasks.In addition,the existing sequence model is difficult to learn the feature representation of medical entities.Therefore,in response to the above problems,here are based on the fusion of knowledge graphs.The use of graph convolutional networks and the variable characteristics of entities to carry out a systematic research on the task of medical entity identification and standardization of online medical consultation texts.Aiming at the problem of insufficient utilization of medical background knowledge,this thesis proposes a named entity recognition model of online medical consultation text that combines background knowledge and attention mechanism.The model extracts the conceptual features of words and candidate knowledge sets from the self-constructed medical knowledge graph,implements joint embedding of the encoded conceptual features and word vectors,and merges the candidate knowledge into the bidirectional Long Short-Term Memory neural network model.Finally,the candidate knowledge set is used to expand the attention mechanism to capture the candidate knowledge and important information in the context again.Aiming at the problem that it is difficult to standardize medical entities,this thesis proposes an online medical consultation text entity standardization model based on graph convolutional networks and entity global features.The model uses a self-defined method to obtain keywords and context from the entity description and context and then obtains text features similar to the denotational semantics.Then,it encodes the global features of the entity through graph convolutional network coding.Besides,this thesis uses the mutual attention mechanism to learn the association between referents and entities.It uses the entity normalization model to introduce keywords further to enhance the mutual attention mechanism.This thesis has carried out entity recognition and standardization on the medical consultation text on the Haodf website to verify the effectiveness of the method in this thesis.Experiments show that compared with the current model using knowledge graphs,the medical entity recognition model in this thesis can improve the performance of the named entity recognition task of online medical inquiry text to a certain extent;compared with the existing models,the medical entity standardization model in this thesis can get better results.