RepOSE : A System for Debugging is-a Structure in Networked Taxonomies

Introduction. Developing ontologies is not an easy task and often the resulting on-tologies are not consistent or complete. Such ontologies, although often useful, lead to problems when used in semantically-enabled applications. Wrong conclusions may be derived or valid conclusions may be missed. RepOSE (Repair of Ontological Structure Environment) tackles the problem of debugging the is-a structure of a fundamental kind of ontologies, i.e. taxonomies. It is a system that supports domain experts in detecting and repairing wrong and missing is-a relations. Using this system we have debugged the ontologies of the Anatomy track of OAEI 2010: MA (2744 concepts and 1807 asserted is-a relations) and NCI-A (3304 concepts and 3761 asserted is-a relations). The debugging took ca 5 hours and resulted in 107 new is-a relations added to MA and 64 to NCI-A, together with 3 asserted is-a relations removed from MA and 12 from NCI-A. System. The input to RepOSE is an ontology network consisting of taxonomies (to be debugged) and correct mappings between the taxonomies. The debugging process consists of the phases of detecting and validating possible defects, and repairing wrong and missing is-a relations. The whole process is driven by the end user (domain expert). At any time during the process, the user can switch between different ontologies, start earlier phases, or switch between the repairing of wrong and missing is-a relations. The process ends when there are no more defects or defect suggestions to deal with. In the current version of RepOSE we have focused on detecting defects using the knowledge inherent in the network. (Other approaches are complementary and can also be used.) RepOSE suggests defects in the form of possibly missing is-a relations which are then validated by a domain expert. This gives us missing and wrong is-a relations. For these defects RepOSE computes repairing actions, i.e. is-a relations to add to and remove from the ontologies such that the missing is-a relations will be derivable from their host ontologies and the wrong is-a relations will not be derivable from the ontology network. To our knowledge, this is the first system that deals with both missing and wrong is-a relations in networked ontologies, and where the repairing of missing is-a relations is more advanced than just adding them to the ontologies. Detection and validation. In RepOSE, the user loads the ontologies and mappings. Then the user can choose an ontology and click Generate Candidate …