Design And Implementation Of Intelligent Medical Assistant Based On Knowledge Graph

The intelligent medical assistant based on the knowledge graph is an auxiliary diagnosis and treatment tool that uses the knowledge graph as the data source.In order to realize the medical assistant,it is necessary to construct a knowledge graph as a data source and an automatic question answering subsystem for interaction.However,there are still some problems in the construction of knowledge graph and automatic question answering subsystem,such as the lack of high-quality data set in Chinese domain,the existing knowledge inference algorithm is difficult to effectively use the topological information in knowledge graph,and the existing semantic understanding model in medical domain is not accurate enough to understand user problems.In view of the above problems,this dissertation studies the related algorithms such as knowledge extraction,knowledge inference and semantic understanding involved in the construction of the system.The main work and innovations are as follows:(1)Aiming at the problem of the lack of high-quality Chinese labeled medical data in medical knowledge extraction tasks,after in-depth research on remote supervised learning,reinforcement learning and attention mechanisms,we proposed the Lattice-LAN model and the RL-Del training algorithm.Compared with the traditional biLSTM-CRF model,Lattice-LAN can make full use of Lattice and Label information,so it has a stronger learning ability for named entity recognition tasks in scenarios with limited data.The RL-Del training algorithm can effectively filter the noise introduced by automatically labeled data in the process of distance supervised learning,so it can achieve better performance in the absence of manually labeled data in the entity relationship extraction task.(2)For medical knowledge inference tasks,the existing methods are difficult to make full use of the rich topological information in the knowledge graph.After in-depth research on neural network technology,knowledge representation and knowledge inference technology,we propose to use the TransE-GCN model for medical knowledge graph inference.Compared with the improved TransE model,this model can use graph convolutional network technology to improve the modeling ability of knowledge graphs.(3)In order to improve the performance of the semantic understanding algorithm in the automatic question answering module,SPF is used to jointly train the two subtasks of intent recognition and slot filling in the semantic parsing task.Under the premise of keeping the structure of the neural network unchanged,this training method can increase the intent recognition F1-measure by about 2%,and the slot filling F1-measure by about 5%,which effectively improves the system’s ability to understand user questions.(4)Based on the proposed algorithm,an intelligent medical assistant based on the knowledge graph is designed and implemented.The test results show that the intelligent assistant can provide friendly interaction,communicate with users through natural language and accurately answer user questions,achieving the design goals.