Study On Automatic Construction And Updating Of Cardiovascular Disease Knowledge Graph

Cardiovascular disease is a group of diseases including coronary heart disease,myocardial infarction,angina,hypertension,arrhythmia,heart failure,and other related illnesses.With changes in people’s lifestyles,the incidence of cardiovascular disease is increasing,becoming an important public health problem worldwide.With the help of artificial intelligence and knowledge graph technology,medical and drug knowledge graphs can be constructed to help doctors understand and master medical knowledge better,and to promote new diagnosis,treatment,prevention,and drug development.Therefore,the application of cardiovascular disease knowledge graphs has broad prospects and can provide powerful support for the prevention,diagnosis,and treatment of cardiovascular disease.However,with the extensive research on cardiovascular diseases and the continuous emergence of research results,traditional knowledge graph construction requires a lot of human effort to carry out tasks such as data cleaning,named entity recognition,relation extraction,conflict discovery,and so on.Therefore,achieving automatic construction and updating of cardiovascular disease knowledge graphs can improve the accuracy and completeness of knowledge graphs,realize timely knowledge updates,and support personalized medical decision-making.To this end,this work proposes an Auto-construction and Self-reflection Framework for Biomedical Knowledge Graph(ASK)for automatic construction and updating of a cardiovascular disease knowledge graph.The framework integrates named entity and relation extraction,link prediction,and conflict resolution techniques based on biomedical text content to achieve efficient,accurate,and reliable 24/7automatic construction and updating of the knowledge graph.The ASK framework consists of three modules: automatic construction,graph update,and self-reflection.The construction module periodically retrieves medical data text from specified sources and extracts entities and relationships using a natural language processing model based on BioBERT.This model increases the automation of knowledge graph construction and reduces the workload of manual data collection and graph construction.In the knowledge graph update module,entities and relationships automatically extracted from the text corpus are integrated into the existing knowledge graph.A conflict resolution algorithm is introduced to detect and resolve conflicts that may arise from new knowledge,improving the accuracy and quality of knowledge graph fusion.Through the self-reflection module,we predict implicit links in the knowledge graph and exclude the possible conflicts caused by implicit links,enriching the content of the knowledge graph,and adjusting and optimizing it accordingly.During conflict resolution,we can detect and resolve issues such as contradictions,errors,and duplicates in the knowledge graph,thus improving the accuracy and completeness of the predicted knowledge in real-time.During our experimental stage,we conducted a series of model and algorithm comparisons to demonstrate the reasons for selecting specific models or algorithms in each module of the framework.We also provided the evolution of the knowledge graph in different modules of the ASK framework.