The activity of identifying and specifying semantic correspondences (e.g., equivalence, subsumption, property mappings) between concepts and relations in two or more ontologies so that data, queries, or inferences can be interpreted consistently across them.

Principle

Alignment produces explicit mappings that translate or relate ontological constructs; these mappings are necessary for automated cross-ontology interpretation, reasoning, or data exchange.

Demonstration

Illustrative scenario → Two biomedical ontologies model a heart attack: Ontology A uses class MyocardialInfarction with property hasOnset; Ontology B uses class HeartAttack with property onsetTime. Recognition → A mapping is declared Equivalence(MyocardialInfarction, HeartAttack) and PropertyMap(hasOnset, onsetTime). Action → A federated query against A is translated to B using the mappings. Consequence → Query returns consistent results from both sources and a reasoner can infer shared class membership.

Misapplication

Treating alignment as matching only lexical labels (string equality) without considering modeling differences (class vs. event, property domains/ranges) leads to incorrect equivalences and invalid inferences.

Consequence

Correct alignment enables integrated queries, aggregated statistics, and cross-source reasoning; incorrect or missing alignments produce inconsistent aggregates, spurious inferences, or missed matches.

Reversal

When ontologies intentionally encode incompatible modeling commitments (e.g., one models events as classes, the other as reified processes) or when provenance must be preserved, full alignment may be infeasible or undesirable; partial or annotated mappings (with provenance and confidence) are then preferable.

Boundary

Clearly within: declaring class and property correspondences between formally modeled ontologies. Boundary case: mapping an ontology class to a term in a controlled vocabulary with weaker formal semantics. Clearly outside: instance-level record linkage or mere schema-level syntactic transformation without semantic correspondence.

Semantic Tension

Expressivity versus tractability — high-fidelity many-to-many mappings preserve meaning but increase computational complexity and ambiguity; strict one-to-one equivalences simplify processing but risk semantic loss.

Synthesis

Ontology alignment formalizes intended semantic equivalences so machines can interoperate, but its utility depends on explicit modeling assumptions, mapping granularity, and documented provenance.