Knowledge graph (KG) represents the real world knowledge in a structured form. KG can be denoted as a set of triples of the form ⟨subject, relation, object⟩. There are three main tasks can be seen in KG, namely:
knowledge graph embedding (KGE), knowledge graph completion (KGC), Knowledge Graph Alignment (KGA). KGE aims to map entities and relations into low-dimensional vectors, which usually regarded as the sub-task in KGC and KGA. The purpose of KGC is to predict new triples on the basis of existing triples, so as to further extend KGs. KGC is usually considered as a link prediction task. KGA aims at finding corresponding nodes or edges referring to the same entity or relationship in different knowledge graphs. KGA, such as cross-lingual knowledge graphs alignment, is useful for constructing more complete and compact KGs.