Extracting information from heterogeneous information sources using ontologically specified target views

摘要

Being deluged by exploding volumes of structured and unstructured data contained in databases, data warehouses, and the global Internet, people have an increasing need for critical information that is expertly extracted and integrated in personalized views. Allowing for the collective efforts of many data and knowledge workers, we offer in this paper a framework for addressing the issues involved. In our proposed framework we assume that a target view is specified ontologically and independently of any of the sources, and we model both the target and all the sources in the same modeling language. Then, for a given target and source we generate a target-to-source mapping, that has the necessary properties to enable us to load target facts from source facts. The mapping generator raises specific issues for a user's consideration, but is endowed with defaults to allow it to run to completion with or without user input. The framework is based on a formal foundation, and we are able to prove that when a source has a valid interpretation, the generated mapping produces a valid interpretation for the part of the target loaded from the source.