Building views over semistructured data sources

The goal of this paper is to present concepts and techniques to build views and organize the information search space of heterogeneous semistructured data sources with respect to expected queries. We formalize the notion of object pattern and of semantic correspondence between object patterns. Object pattern analysis is ontology-driven and leads to the construction of reconciled views of the sources, called global object patterns. Global object patterns mediate between heterogeneous terminology and structure of data in different sources and are used for query formulation and data extraction.