Research On The Construction And Application Of Diabetes Knowledge Graph

With the improvement of my country’s economy and residents’ consumption level,people’s living patterns and dietary structure have changed,resulting in an increase in the number of diabetic people,and the younger population tends to be prominent.Therefore,the research focus of this paper is to explore how to use digital technology to help diabetic users obtain diabetes prevention and treatment knowledge more easily and efficiently through the Internet platform,further promote the development of the "Internet +" medical model,improve the management level of diabetes in my country,and improve people’s health status.As traditional search engines often return redundant information,users need to spend a lot of time and effort to sift through it.The question answering system pays attention to understanding the user’s intentions,and can directly feedback the answer according to the user’s question.The knowledge graph integrates multi-source data and provides a structured knowledge base to help the question answering system better understand user queries and provide accurate answers.Therefore,this paper studies and implements a diabetes question answering system based on knowledge graph.In order to enhance the system’s question parsing ability,this paper adopts a named entity recognition model based on deep learning.After testing,the system verified the application of knowledge graph in diabetes question answering system.The main work of this paper is as follows:(1)Construct a knowledge graph in the field of diabetes.Crawl diabetes-related data from pharmaceutical websites,perform knowledge extraction after cleaning,and perform multi-source data knowledge fusion with the Ruijin Hospital diabetes dataset to improve the integrity and accuracy of the knowledge graph.Then the data after knowledge fusion is stored in the Neo4 j graph database to complete the construction of the knowledge graph in the field of diabetes.(2)Research on algorithms for named entity recognition.Firstly,the diabetes question-and-answer data is crawled from major pharmaceutical websites as a data set for model training,and word segmentation and entity labeling are performed on the data set,and then a deep learning method that embeds the BiLSTM-CRF model on the basis of the BERT pre-training model is studied and constructed,for named entity recognition tasks.Then through the comparative experiment with the other two models,the effect of the model is verified.(3)Construction of diabetes question answering system based on knowledge graph.Based on the previous two research works,functions such as the display of diabetes knowledge graph,the retrieval of disease relationship,and knowledge question-answering were designed and implemented.The main implementation steps of the automatic question answering system are explained one by one,and its operation effect is shown,and finally the overall function test of the question answering system is done.The question answering system implemented in this paper has been tested and the accuracy rate of answering diabetes-related questions has reached more than 80%.It can help users obtain diabetes-related knowledge easily and effectively without leaving home.The concise and clear system interface design and easy-to-use input methods provide users with a good experience.