RAHC: Rule-Augmented Hyper-Relational Knowledge Graph Completion for Materials Synthesis
| Source: arXiv AI
Tags: RAHC, Hyper-Relational Knowledge Graphs, Knowledge Graph Completion, Materials Synthesis, MAYPL, ILPC'22
RAHC fills in missing materials-synthesis conditions by combining a hyper-relational knowledge graph with a traceable rule catalog, gaining up to 15.6% MRR over MAYPL on semi-inductive completion.
Details
Experimental records in materials synthesis are often incomplete, and recovering a missing condition depends on the target material and the context of a specific experiment. Hyper-relational knowledge graphs capture that context through qualifiers, but existing completion methods keep domain rules implicit. RAHC builds a rule catalog from training records and retrieved literature, then retrieves rules for each query-candidate pair. Hard rules filter invalid candidates and soft rules guide ranking. RAHC-G weights rules by global reliability, while RAHC-Q also learns query-specific relevance; gated fusion merges graph structure and rule evidence. Across five materials-synthesis datasets, the method improves single-slot and joint multi-slot completion, and it also helps on the entity-disjoint ILPC'22 Small benchmark, suggesting transfer to fully inductive KG completion.