Curating the Specificity of Ontological Descriptions under Ontology Evolution

Abstract

Semantic Web Ontologies are not static, but evolve, however, this evolution usually happens independently of the ontological instance descriptions (usually referred as “metadata”) which are stored in the various Metadata Repositories or Knowledge Bases. Nevertheless, it is a common practice for a MR/KB to periodically update its ontologies to their latest versions by “migrating” the available instance descriptions to the latest ontology versions. Such migrations incur gaps regarding the specificity of the migrated metadata, i.e. inability to distinguish those descriptions that should be reexamined (for possible specialization as consequence of the migration) from those for which no reexamination is justified. Consequently, there is a need for principles, techniques, and tools for managing the uncertainty incurred by such migrations, specifically techniques for (a) identifying automatically the descriptions that are candidates for specialization, (b) computing, ranking and recommending possible specializations, and (c) flexible interactive techniques for updating the available descriptions (and their candidate specializations), after the user (curator of the repository) accepts/rejects such recommendations. This problem is especially important for curated knowledge bases which have increased quality requirements (as in E-science). In this paper, we formalize the problem and propose principles, rules, and algorithms for realizing the aforementioned scenario, for the RDF/S framework. Finally, we report experimental results demonstrating the feasibility of the approach.