A COVID-19 knowledge model integrating traditional Chinese and western medicine knowledge and an automatic domain knowledge discovery method based on Natural Language Processing (NLP) are proposed, and the semantic knowledge is organized and associated from the hierarchical structure, subject classification, relation types and other dimensions. Additionally, the COVID-19 literature, diagnosis and treatment regimens and the new media data are collected for knowledge discovery and manual annotation, and the COVID-19 knowledge graph is built and realized for visual presentation and online retrieval, so as to support the well organization and effective utilization of biomedical literature resources.