Ontology Alignment

Ontology alignment is the process of establishing correspondences between the concepts and relationships of two or more distinct ontologies. In multi-agent systems, alignment is not a preprocessing step but an ongoing negotiation: agents with different internal models must discover shared structure through interaction, not merely map pre-defined categories.\n\nThe problem is formally equivalent to the graph matching problem in computer science and to the translation problem in linguistics. But the agent communication case adds a dynamic dimension: the ontologies themselves evolve as agents learn, making alignment a moving target rather than a static mapping.\n\nThe Semantic Web assumed that ontology alignment could be achieved through formal specification. The failure of that assumption in open systems has shifted attention to interactive alignment — agents that learn to communicate by communicating, building shared meaning through trial and error rather than pre-negotiated dictionaries.\n\n