Reasoning-supported interactive revision of knowledge bases

Abstract

Quality control is an essential task within ontology development projects, especially when the knowledge formalization is partially automatized. We propose a method for integrating newly acquired, possibly low-quality axioms into an existing ontology after their manual inspection; based on the decision whether the axiom is desired or not, several of the yet unevaluated axioms are evaluated automatically. Since the evaluation order can significantly increase the amount of automatization, we further propose the notion of axiom impact. Finally, we introduce decision spaces as structures to efficiently compute the axiom impact and the implicit evaluation decisions. Compared to a naïve implementation, this reduces the number of costly reasoning operations on average by 75%.