Forgetting, and the dual task of computing uniform interpolants, minimally restricts an ontology to a given set of concept and role symbols in such a way that consequences involving these symbols are preserved. This makes forgetting/uniform interpolation a useful tool for module extraction, ontology analysis, ontology evolution and information hiding. Given the importance for all these applications, forgetting/uniform interpolation has recently gained a lot of attention in the description logic and knowledge representation literature.
Previous work devised methods for simpler Horn description logics (such as DL-Lite and EL) and for concept symbol elimination of ALC ontologies.
In recent work we have developed a new method for computing uniform interpolants for a large number of more expressive description logics with the advantage that it can eliminate both concept and role symbols and can handle ontologies that include ABox statements (facts).
Experiments on a large corpus of ontologies from real-world applications show that our method is practical and widely applicable.
This is joint work with Patrick Koopmann.