Study On Representation Learning And Application Of Medical Knowledge Graph

In the era of big data,with the rapid growth of data,an urgent issue is how to store and use these data reasonably.At present,the medical record data still adopts the traditional model of simple storage in the medical field.Although it has evolved into the electronic medical record,it is still complex and has not been effectively used.Since 2012,Google has proposed the concept of knowledge graphs to improve the search efficiency.The purpose of this paper is to explore the application system of building knowledge map,analyze the existing application needs,improve the function of knowledge map system,and integrate the existing knowledge map application technology into the system of this design.After analyzing the current situation at home and abroad,I found that it is an urgent need to build a knowledge map in the medical field,but the professional knowledge in medicine hinders the research progress of researchers.And,there is no open-source medical data with annotation in China.Therefore,there are still serious deficiencies in the study of Chinese medical knowledge map.In the process of building the knowledge map system,data acquisition and knowledge extraction become the first step to solve the problem.After the knowledge triples are extracted,the knowledge graph is represented and learned,and applied to the system to realize the knowledge link prediction task,and then the knowledge question and answer module is added to form a complete knowledge graph system.The main work of this paper: collected relevant data sets,screened and extracted part of the data in these data sets,constructed a new data set and manually label a data set;compared the existing models and select a knowledge extraction model;compared the models commonly used in knowledge representation learning at this stage,these models have been improved by optimizing model input,and after experimental comparison,the representation learning models that perform better are selected;By consulting materials and analyzing other papers,comparing different knowledge graph storage schemes,choosing a storage method that is most suitable for the system to store the knowledge graph in the system,which can not only speed up the search,but also provide a visual display;The research content includes a variety of knowledge graph application modules,including knowledge link prediction module,knowledge visual display module,and knowledge question answering module,to build a relatively complete knowledge graph application system.