| How to quickly and efficiently guide patients to register and seek treatment is an urgent problem for large general hospitals in the new era.Artificial intelligence technology based on deep learning has become a hot research topic at this stage as a tool to assist medical decision making and applied to clinical work.The intelligent triage system studied in thesis analyzes patients’ complaints by establishing a triage recommendation model,realizes the function of recommending departments for patients,and improves the accuracy of the triage model by designing a symptom weighting algorithm.The main research contents of thesis are as follows:1)A joint extraction model of entity and relations based on sequence annotation for Chinese medical is designed.To address the problem of relationship overlap between triads,thesis proposes an improved joint extraction strategy,decomposes the extraction task into two related subtasks,and builds a joint extraction model of entity relationship based on Ro BERTa-GRU-Bi LSTM.In the c Me IE extraction task,the experimental results show that the method in thesis exhibits better extraction results compared with four extraction methods based on other pre-trained language models.2)A triage recommendation algorithm based on chief complaint information is proposed.To meet the patient’s triage demands and achieve accurate department recommendation service,thesis proposes a Bi LSTM triage recommendation model with the addition of Attention mechanism,completes the case database augmentation and the specification processing of department name,designs a symptom weighting algorithm,and incorporates medical history features as model inputs.The experimental results verify that compared with the mainstream Bi LSTM triage algorithm based on Ro BERTa,the evaluation value of the triage performance of this method is improved by 1.43%,and achieved a more effective department classification work.3)Implemented an intelligent triage system with privacy protection function.To address the problem of privacy leakage of patients’ conditions faced in the process of enjoying medical services,thesis proposes anonymous medical consultation and risk prediction privacy security access control strategies based on sensitive mechanisms.Experimental results show that compared with k-anonymity and information entropy l-diversity models,the model in thesis achieves higher equivalence class information entropy,while effectively mitigating data loss.Thesis completes the design and implementation of an intelligent triage system to realize the function of providing online triage consultation service for patients.The intelligent triage system designed in thesis can provide more accurate department recommendation services for patients,provide patients with a convenient online triage service window,achieve stronger complaint understanding,and improve triage services for patients. |
