Learning data-consistent mappings from a relational database to an ontology

Open Interop Workshop on Enterprise Modelling and Ontologies for Interoperability

Benjamin Habegger

Others  :  http://CEUR-WS.org/Vol-200/06.pdf
PID  :  3962

来源: CEUR

【 摘 要 】

Determining how to map information between different representational models is an essential task of semantic integration. Up to now, most research in this field has concentrated on the problem of find- ing pairs of elements which could have similar semantics (ie. finding matches). From the work we have seen, it is not clear that any guarantees can be given that the obtained matches respect the intended semantics of the target model. We also argue that matches are not sufficient to clearly define mappings which respect the intended interpretation of the target model. In this paper, we define mapping consistency with respect to an intended interpretation and derive a notion of data-consistency, a necessary condition for consistency. From this, we propose a relational learning algorithm to construct data-consistent mappings.