Bisimulation-based concept learning for information systems in description logics

摘要: In description logic-based information systems, objects are described not only by attributes but also binary relations between them. This work studies concept learning in such systems. It extends the bisimulation-based method of Nguyen and Szalas (Rough sets intelligent Springer, Berlin, pp 517–543, 2013). We take as basic elements language. Each attribute may be discrete or numeric. A Boolean is treated a name. approach more general suitable for practical systems based on logic than one As further extensions, we allow data roles constructors “functionality” “unqualified number restrictions”. formulate prove an important theorem selectors. present domain partitioning gain that has been used our implementation method. Apart from selectors simple selectors, introduce new kind called extended The evaluation results show valuable support it significantly.