One of the most important elements of culture is painting, which reserves ethnic and national immaterial relics. However, Chinese Web is currently short of a specialized knowledge graph on museums and it still crave accesses to effectually assimilate extensive types of discrete knowledge for applications. To facilitate museum knowledge sharing, we propose a painting knowledge graph, namely PaintKG, utilizing Bi-LSTM with CRF layer which obviously rises the F-l measure to recognize and extract knowledge and relations of different types of paintings and their painters from existing encyclopedia and unstructured Web text data and physically restore them in neo4j as graph data eventually. What’s more, we depict and demonstrate typical scenarios of PaintKG and implement them by real-world applications, such as painting recommendation, painter entity association.