Medical Knowledge Graph to Promote Rational Drug Use: Model Development and Performance Evaluation

Abstract

Abstract Knowledge Graph (KG) has been proven effective in representing and modeling structured information, especially in the medical domain. However, obtaining structured medical information usually depends on the manual processing of medical experts. Meanwhile, the construction of Medical Knowledge Graph (MKG) remains a crucial problem in medical informatization. This work presents a novel method for constructing MKGto drive the application of Rational Drug Use (RDU). We first collect and preprocess the corpora from various types of resources, and then develop a medical ontology via studying the concepts in RDUdomain, authoritative books and drug instructions. Based on the medical ontology, we formulate a scheme to annotate the corpora and construct the dataset for extracting entities and relations. We utilize two mechanisms to extract entities and relations respectively. The former is based on deep learning, while the latter is the rule-based method. In the last stage, we disambiguate and standardize the results of entity relation extraction to construct and enrich the MKG. The experimental results verify the effectiveness of the proposed methods.

Keywords

• Rational Drug Use
• Medical Knowledge Graph
• Named Entity Recognition
• Relation Extraction