Integration integrity for multigranular data

Abstract

When data from several source schemata are to be integrated, it is essential that the resulting data in the global schema be consistent. This problem has been studied extensively for the monogranular case, in which all domains are flat. However, data involving spatial and/or temporal attributes are often represented at different levels of granularity in different source schemata. In this work, the beginnings of a framework for addressing data integration in multigranular contexts are developed. The contribution is twofold. First, a model of multigranular attributes which is based upon partial orders which are augmented with partial lattice-like operations is developed. These operations are specifically designed to model the kind of dependencies which occur in multigranular modelling, particularly in the presence of aggregation operations. Second, the notion of a thematic multigranular comparison dependency, generalizing ordinary functional and order dependencies but specifically designed to model the kinds of functional and order dependencies which arise in the multigranular context, is developed.