Research On Construction Of Medical Knowledge Graph And Its Application

In recent years,Internet-related technologies are developing rapidly,and people's lifestyles are changing with the development of technologies.Medical and health issues have always been one of the most concerned issues in people's lives.On one hand,naturally,there turn out to be more and more websites on the medical and health aspects,and medical information is becoming more and more abundant.On the other hand,with the development of electronic devices,the way that many hospitals record patient information has also been updated from on the traditional paper medical records to through the use of computer systems to store electronic medical records.The variety of information is too much and complex.It is difficult for people to catch the information they really need from heaps of medical information.Knowledge graphs provide an excellent solution for the management of knowledge.Knowledge in the medical field is specialized,compl icated and confused.If information in the medical field can be organized in the form of a knowledge graph,it will be of great help to the further application of medical knowledge in people's lives.The construction of knowledge graphs in the medical field is an issue which people urgently want to solve.However,the professionalism of knowledge in the medical field brings a lot of inconvenience to the construction of the knowledge graph.First of all,there are not many labeled data containing medical expertise,and there is not enough data available for direct use.Unlabeled data contain s more medical expertise but is not effectively utilized.Second,some concepts of certain words in the medical field are different from that in the general domain.It is not good to directly refer the way of constructing knowledge graphs in the general domain to the medical field.In addition,most knowledge graphs related research focuses on the relationship between one entity and another.However,for medical treatment,the attribute and the value of the medical attribute plays an important role in the analysis of medical cases,it will be better if the relationship of <entity,attribute,value> triples are well-organized in the knowledge graphs.This paper discusses the method of extracting attribute knowledge and value of attribute knowledge in medical domain from unlabeled data,and solves the problem that it is hard to process complicated medical data automatically.In order to overcome the trouble of natural language processing application in medical domain,this paper aims to explore a method of constructing knowledge graphs in medical field,and studies and implements a semi-supervised method for mining knowledge from medical information.Through using Bootstrapping algorithm and conditional random field(CRF)model,we constructed the dictionary of medical field entities,then designed a method to mine the data needed for constructing knowledge graphs;tried to improve the information extraction performance with the method of training word vectors in medical domain;studied the method of combining different knowledge graph data;and tried to apply the knowledge graphs to solve practical problems.