Ontology Mediation

Mediation strategies are categorized by when the semantic reconciliation occurs. Design-time mediation involves creating a static, global schema (a merged ontology) that all data sources are mapped to before any queries are executed. Query-time mediation uses a virtual approach, where user queries against a unified ontology are dynamically rewritten into source-specific sub-queries using declarative mappings, as seen in Ontology-Based Data Access (OBDA) systems.

A foundational framework for classifying semantic mismatches that must be resolved during mediation. It identifies conflicts at three levels:

• Model Mismatches: Differences in modeling paradigms (e.g., relational vs. graph).
• Meta-Model Mismatches: Differences in schema constructs (e.g., class vs. property).
• Model-Instance Mismatches: Conflicts between the schema definition and the actual data population. Understanding these layers is critical for selecting the correct alignment and merging strategy.

A query-time mediation strategy where the global unified ontology is defined as a view over the local data sources. Each class or property in the global schema is associated with an explicit query (e.g., in SPARQL or SQL) that specifies exactly how to retrieve its instances from the underlying sources. This approach simplifies query rewriting but can make adding new data sources difficult, as the global view must be manually updated.

An alternative query-time strategy where each local data source is defined as a view over the global ontology. The mediation engine must perform complex reasoning to reformulate a global query into queries on these local views. While adding new sources is trivial (just describe the new view), query reformulation is computationally intensive. This is the foundational paradigm behind Ontology-Based Data Access (OBDA) using R2RML mappings.

Mediation involves two distinct technical operations:

• Ontology Alignment: The process of identifying correspondences (mappings) between entities in different ontologies, typically using the Alignment API format to represent owl:sameAs or rdfs:subClassOf links.
• Ontology Merging: The process of taking the source ontologies and the generated alignment to create a single, unified, and coherent ontology. This step often requires alignment repair to resolve logical inconsistencies introduced by the mappings.

The core computational challenge of query-time mediation. A user's SPARQL query, written against the global ontology's vocabulary, must be algorithmically rewritten into a union of queries against the physical schemas of the underlying data sources. This process must be semantically preserving, meaning the rewritten query must return exactly the same answers that the original query would if all data were materialized under the global ontology, a property verified through SPARQL entailment regimes.

The computational process of determining correspondences between heterogeneous ontologies to achieve semantic interoperability. Also known as ontology matching, it identifies equivalence, subsumption, and disjointness relations between entities. Modern systems combine lexical matchers, structural analyzers, and neural embedding models to generate candidate mappings with confidence scores.

The divergence in meaning or interpretation of data across different schemas, representing the primary obstacle to automated knowledge graph interlinking. Manifests in four forms:

• Terminological heterogeneity: Different names for the same concept
• Conceptual heterogeneity: Different modeling granularity or coverage
• Syntactic heterogeneity: Different representation formats
• Semiotic heterogeneity: Different interpretations by human communities

The post-matching process of detecting and removing incoherent correspondences from a generated alignment to restore logical satisfiability. When mappings introduce contradictions—such as violating the Conservativity Principle by creating unintended subclass relationships—repair algorithms prune the minimal set of correspondences needed to restore consistency while maximizing alignment coverage.

A virtual data integration paradigm that uses an ontology as a high-level conceptual schema to mediate queries across heterogeneous data sources. OBDA systems employ R2RML mappings to translate SPARQL queries into native SQL, enabling transparent access without physical data consolidation. The ontology serves as the mediated schema that users query, while the system handles reformulation to underlying sources.

An algorithmic solution applied to ontology matching that finds a stable one-to-one mapping between two sets of entities based on mutual preference scores. Each entity ranks potential matches by similarity, and the algorithm produces a pairing where no two entities would prefer each other over their assigned partners. This optimizes global alignment cardinality and prevents conflicting one-to-many mappings.