iTrails: Pay-as-you-go Information Integration in Dataspaces

Overview In their work, the authors consider a novel model of heterogeneous data organization called dataspaces. It is aimed to improve the two most popular approaches used in data integration: the schema first approach (SFA) and the no schema approach (NSA). The SFA approach requires semantically integrated view over integrated data sources. Query languages generally have formal semantics, however maintaining the mediated schema here is generally costly. The NSA approach requires no semantically integrated view and considers sources as XML-like documents. Since no mediated schema is present, no semantic mappings between schemas are required. At the same time, query languages here a limited to XPath-like languages and keyword search. The model considered by the authors is in between there two approaches which gives it advantages (and some disadvantages) of both. It starts with NSA and gradually approaches SFA in a pay-as-you-go manner by providing semantic hints (or, trails) to query engine.