DHGE: Dual-View Hyper-Relational Knowledge Graph Embedding for Link Prediction and Entity Typing

In the field of representation learning on knowledge graphs (KGs), a hyper-relational fact consists main triple and several auxiliary attribute-value descriptions, which is considered more comprehensive specific than triple-based fact. However, currently available KG embedding methods in single view are limited application because they weaken hierarchical structure that represents affiliation between entities. To overcome this limitation, we propose dual-view (DH-KG) contains instance for entities ontology concepts abstracted hierarchically from This paper defines link prediction entity typing tasks DH-KG first time constructs two datasets, JW44K-6K, extracted Wikidata, HTDM based medical data. Furthermore, DHGE, model GRAN encoders, HGNNs, joint learning. DHGE outperforms baseline models DH-KG, according to experimental results. Finally, provide an example how technology can be used treat hypertension. Our new datasets publicly available.