Towards an Error-Tolerant Construction of $\mathcal{EL}^\bot$-Ontologies from Data Using Formal Concept Analysis

Abstract

In the work of Baader and Distel, a method has been proposed to axiomatize all general concept inclusions (GCIs) expressible in the description logic $\mathcal{EL}^\bot$ and valid in a given interpretation $\mathcal{I}$. This provides us with an effective method to learn $\mathcal{EL}^\bot$-ontologies from interpretations. In this work, we want to extend this approach in the direction of handling errors, which might be present in the data-set. We shall do so by not only considering valid GCIs but also those whose confidence is above a given threshold c. We shall give the necessary definitions and show some first results on the axiomatization of all GCIs with confidence at least c. Finally, we shall provide some experimental evidence based on real-world data that supports our approach.