NQE: N-ary Query Embedding for Complex Query Answering over Hyper-Relational Knowledge Graphs

Complex query answering (CQA) is an essential task for multi-hop and logical reasoning on knowledge graphs (KGs). Currently, most approaches are limited to queries among binary relational facts pay less attention n-ary (n≥2) containing more than two entities, which prevalent in the real world. Moreover, previous CQA methods can only make predictions a few given types of cannot be flexibly extended complex queries, significantly limits their applications. To overcome these challenges, this work, we propose novel N-ary Query Embedding (NQE) model over hyper-relational (HKGs), include massive facts. The NQE utilizes dual-heterogeneous Transformer encoder fuzzy logic theory satisfy all FOL including existential quantifiers (∃), conjunction (∧), disjunction (∨), negation (¬). We also parallel processing algorithm that train or predict arbitrary single batch, regardless kind each query, with good flexibility extensibility. In addition, generate new dataset WD50K-NFOL, diverse WD50K. Experimental results WD50K-NFOL other standard datasets show state-of-the-art method HKGs generalization capability. Our code publicly available.