Supporting manual mapping revision using logical reasoning
Christian Meilicke, Heiner Stuckenschmidt, Andrei Tamilin · MADOC (University of Mannheim) · 2008
Finding correct semantic correspondences between ontologies is one of the most challenging problems in the area of semantic web technologies.Experiences with benchmarking matching systems revealed that even the manual revision of automatically generated mappings is a very difficult problem because it has to take the semantics of the ontologies as well as interactions between correspondences into account.In this paper, we propose methods for supporting human experts in the task of revising automatically created mappings.In particular, we present non-standard reasoning methods for detecting and propagating implications of expert decisions on the correctness of a mapping.We show that the use of these reasoning methods significantly reduces the effort of mapping revision in terms of the number of decisions that have to be made by the expert. MotivationA common way of integrating different ontologies describing the same or largely overlapping domains is to use formal representations of semantic correspondences between their concepts and relations -also referred to as 'ontology mappings'.Recently, a number of automatic and semi-automatic tools for generating hypotheses about semantic correspondences have been developed (see (Euzenat & Shvaiko 2007) for an overview).The results of these tools, however, often contain a significant amount of errors caused by the use of general heuristics that are bound to fail in certain situations.Due to this fact, a manual revision of the mappings created by a matching system is often inevitable.Revising mappings is a very complex and difficult problem even for experts in the area.We can identify two sources of complexity:• The correctness of mappings depends on the semantics of the ontologies.Therefore, mapping revision requires to completely consider the ontologies linked by the mapping.This makes some form of logical reasoning indispensable which is almost impossible to do manually due to the size and complexity of the ontologies.• Individual decisions about the correctness of a suggested semantic relation can have an influence on past and fu-