ID.6. Defining Flexible Mappings to Empower Dynamic Information Discovery

The success of data warehousing has given rise to new highly dynamic and flexible analytical settings. In such settings, the end-users are supposed to pose analytical queries over disparate data sources in right time. However, identifying the needed subset of source data to answer the end-user analytical needs is of paramount importance.

In traditional data warehousing settings, the sources are mainly relational and static, and this process could be automated up to some extent by inferring a mapping between data sources and data warehouse schemas. Nevertheless, even in such controlled settings, blindly exploring these mappings would be a huge mistake, resulting in useless (maybe erroneous) analysis perspectives of no relevance for the end-user. Thus, semantics must come into play and lead the search. In terms of popular semantic constructs from the Semantic Web such as ontologies, this can be done by means of reasoning.

There are two main trends to define mappings between the data sources and the domain of discussion (usually captured in terms of the integration layer schema): namely GaV (Global as View) and LaV (Local as View). The former result in cheaper computations, but it is too rigid. Oppositely, the latter is much more flexible and adequate for analytical settings, but results to be expensive (easily unfeasible) in the general case.

In this challenge we aim at further exploring how to couple reasoning with LaV mappings. An interesting approach would be limiting the reasoning techniques to the multidimensional setting and, in this way, balance the expressivity and feasibility of reasoning over LaV mappings. As result, we would be able to deal and integrate different kinds of data sources within the analytical setting at hand, even some poorly structured like those we usually find in the Web.

Main Advisor at Universitat Politècnica de Catalunya (UPC)
Co-advisor at Université Libre de Bruxelles (ULB)