Flexible Mapping for Retrieval from Federated Schemas with Modelling Variations

Retrieving data from diverse information sources is a problem faced by many organizations. Many of these information sources are databases where diierent schemas may represent the same basic concept in diierent ways. The problem that leads to this semantic heterogeneity is termed modelling variation. Current approaches limit the retrieval of related information from multiple sources, because they do not support a wide range of modelling variations. We propose to work with an abstract schema, represented as an Entity-Relationship Diagram, to represent information of interest to a user or a group of users. The abstract schema may have been created in response to a set of related user requests or queries and is used for retrieval only. Database administrators of individual schemas provide mappings that relate constructs appearing in the abstract schema to constructs in the individual schemas. The abstract schema is populated , by issuing queries against the individual schemas. The contribution of this research is the deenition of the mappings that handle a wider variety of modelling variations between the abstract schema and the individual schemas than previous approaches. These mappings can then be used to populate the abstract schema with valid data from underlying information sources.