An approach to the semi-automatic generation of mediator specifications

Abstract

Mediator architectures have become popular for systems that aim at providing transparent access to heterogeneous information sources. Similar to a view on a database, a mediator answers queries posed against the common domain schema by executing one or more queries against its underlying information sources and mapping the results back into the common schema. The mapping is usually specified declaratively by a set of rules, where the rule heads define queries against a common domain schema and the rule bodies define their implementations in terms of queries against one or more source schemas. In this paper a mechanism is presented that supports finding these mappings, a task that requires a fair amount of knowledge about the underlying schemas and that involves a lot of effort, thus becoming a major obstacle to scalability.