Reducing the Cost of Validating Mapping Compositions by Exploiting Semantic Relationships
Defining and composing mappings are fundamental operations required in any data sharing architecture (e.g. data warehouse, data integration). Mapping composition is used to generate new mappings from existing ones and is useful when no direct mapping is available. The complexity of mapping composition depends on the amount of syntactic and semantic information in the mapping. The composition of mappings has proven to be inefficient to compute in many situations unless the mappings are simplified to binary relationships that represent “similarity” between concepts. Our contribution is an algorithm for composing metadata mappings that capture explicit semantics in terms of binary relationships. Our approach allows the hard cases of mapping composition to be detected and semi-automatically resolved, and thus reduces the manual effort required during composition. We demonstrate how the mapping composition algorithm is used to produce a direct mapping between schemas from independently produced schema-to-ontology mappings. An experimental evaluation shows that composing semantic mappings results in a more accurate composition result compared to composing mappings as morphisms.
mapping, composition, integration, model management, semantics, metadata
Date of this Version