INFDok - Dokumentenvolltextserver des Fachbereichs Informatik

Kurzfassung auf Englisch:

In commonality-based information retrieval, the commonalities of user-specified examples of desired information are used for information retrieval. As described in previous literature, such a retrieval mechanism can be built using a combination of description logic reasoning services which makes this retrieval technique an interesting research topic in the field of description logic applications. However, as will be shown in this article, the existing technique does not always yield a desirable retrieval result. Therefore, we present a theoretical framework useful for commonality-based information retrieval and other description logic applications. Part of this framework is a formal definition of the notion of ABox subsumption using the standard semantics for ABoxes and an algorithm for deciding this problem. Based on ABox subsumption, we will present an operation for determining the least common subsuming ABox of a set of ABoxes w.r.t. the expressive description logics ALENR and ALQ. As a by-product, an algorithm for computing the most specific concept of individuals occurring in ABoxes w.r.t. the two mentioned description logics will be developed. We present soundness, completeness, and complexity results and show how the developed reasoning services can be used for a variant of commonality-based information retrieval which we call commonality-based ABox retrieval.