AgentMap: Joint Equivalence and Subsumption Discovery for Ontology Matching

2026-07-29

AI Summary

This work addresses the limitation of existing ontology matching methods, which typically support only equivalence or subsumption relations but struggle to handle both simultaneously. To bridge this gap, we introduce the hybrid ontology matching task and propose AgentMap, a multi-agent framework grounded in large language models. AgentMap leverages semantic retrieval, hierarchical traversal, and collaborative reasoning to automatically identify either equivalent concepts or the most specific superconcepts for source ontology terms within a target ontology. Our approach is the first to unify the discovery of equivalence and subsumption correspondences through a coordinated multi-agent mechanism and introduces the first benchmark dataset for hybrid ontology matching. Experimental results demonstrate that AgentMap achieves superior performance in hybrid settings and significantly outperforms current baselines on both pure equivalence and pure subsumption tasks.

Abstract

Ontology matching (OM) has traditionally been formulated as either equivalence discovery or subsumption matching. The existing OM systems identify only one type of semantic correspondence and cannot simultaneously discover equivalence and subsumption mappings. In this paper, we introduce Hybrid Ontology Matching (HOM), a new OM task that unifies equivalence and subsumption discovery, and accordingly propose a Large Language Model (LLM)-based multi-agent OM framework AgentMap that is implemented by a series of interdependent semantic decisions. Given a concept in the source ontology, AgentMap integrates semantic retrieval, hierarchical search, and collaborative multi-agent LLM reasoning to progressively explore the target ontology, identifying either the equivalent concept, if one exists, or the most fine-grained subsumer. We further extend four OM datasets for a HOM benchmark and evaluate AgentMap under hybrid, equivalence-only, and subsumption-only settings. Experimental results show that AgentMap achieves promising performance on the hybrid setting, and at the same time outperforms equivalence matching and subsumption matching baselines on the equivalence-only and subsumption-only settings, respectively.

Problem

• Ontology Matching
• Equivalence Discovery
• Subsumption Matching
• Hybrid Ontology Matching

Innovation

• Hybrid Ontology Matching
• Multi-agent LLM
• Equivalence Discovery
• Subsumption Matching
• Semantic Reasoning