AgentMap: Joint Equivalence and Subsumption Discovery for Ontology Matching
AgentMap treats ontology matching as one combined search for both equivalent concepts and finer-grained subsumers.
The paper defines a new Hybrid Ontology Matching task to cover both correspondence types in one framework. AgentMap uses semantic retrieval, hierarchical search, and collaborative multi-agent LLM reasoning to explore a target ontology from a given source concept. The authors extend four datasets into a HOM benchmark and test hybrid, equivalence-only, and subsumption-only settings. They report promising hybrid results and stronger performance than relevant baselines in the single-task settings. ArXiv · AI/CL/LG's note
The paper defines a new Hybrid Ontology Matching task to cover both correspondence types in one framework. AgentMap uses semantic retrieval, hierarchical search, and collaborative multi-agent LLM reasoning to explore a target ontology from a given source concept. The authors extend four datasets into a HOM benchmark and test hybrid, equivalence-only, and subsumption-only settings. They report promising hybrid results and stronger performance than relevant baselines in the single-task settings. ArXiv · AI/CL/LG's note
score 4