Hitting the Sweetspot: Economic Rewriting of Knowledge Bases
- Speaker: Prof Birte Glimm (University of Ulm)
- Host: Konstantin Korovin
- 25th March 2013 at 14:00 in Lecture Theatre 1.3
Given a knowledge base over some signature and a subset of "relevant" terms from this signature, the task of knowledge base extraction is the computation of a knowledge base that contains as little irrelevant information as possible, and, at the same time, contains all information about the relevant terms. Three conflicting requirements arise in this context: the size of the extracted knowledge base, the size of its signature and the syntactic similarity of the extracted knowledge base with the original one. Minimal module extraction and uniform interpolation assign an absolute priority to one of these requirements, thereby limiting the possibilities to influence the other two. In this talk I present a novel technique for knowledge base extraction in the Description Logic EL that does not require such an extreme prioritization. Furthermore, a tractable rewriting approach is presented and an empirical study comparing the technique with existing approaches shows encouraging results.