Intelligent Hospital Guidance System Based on Multi-Round Conversation

Registering a wrong hospital department is common when patients use on-line registering systems. Currently, there are some systems in practice. However, patients are unable to choose the best department due to different names and authorities of hospitals. To help solve the problem, we build a symptom-disease-disciplinary knowledge graph to recommend appropriate departments for patients. We obtain real disease-disciplinary information based on the regional health platform electronic health records (EHRs). Besides, we synthesize the symptom-disease relationship between ICD codes and medical encyclopedia websites. To further help the system predict the diseases based on patients' complaints, we update the weights of diseases through patients' choices in multi-round conversations. Experimental results show that the accuracy of final prediction is up to 92%.