Modeling heterogeneous data in dataspace

Recently, dataspace has been proposed as a new architecture in the evolution of information integration. However, how to fulfill the vision of dataspace, i.e. pay-as-you-go integration, is still a suspensive issue, and the first challenge is data modeling. In this paper, we present a flexible data model called triple model which is similar to RDF but more suitable to dataspace applications. Information in arbitrary data model can be uniformly represented by triple model in dataspace. According to pre-defined rules, each information item would be decomposed into a bunch of structures called triple, and all triples are loosely connected to form a unified view without reconciling semantic heterogeneity upfront. Then users are enabled to query or search in a uniform way over all heterogeneous information sources. Further, when users demand, existing dataspace can be incrementally enhanced by tighter integrations.