Tractable Feature Generation Through Description Logics with Value and Number Restrictions [chapter]

Nicola Fanizzi, Luigi Iannone, Nicola Di Mauro, Floriana Esposito
2006 Lecture Notes in Computer Science  
In the line of a feature generation paradigm based on relational concept descriptions, we extend the applicability to other languages of the Description Logics family endowed with specific language constructors that do not have a counterpart in the standard relational representations, such as clausal logics. We show that the adoption of an enhanced language does not increase the complexity of feature generation, since the process is still tractable. Moreover this can be considered as a
more » ... tion for future employment of even more expressive languages from the Description Logics family. Person(Tom), Male(Tom)} BK descriptions for feature generation T BK Single ≡ Person ≤ 0.isMarriedTo; Mother ≡ ¬Male ∀Parent.Person ≥ 1.Parent; GrandParent ≡ Person ∀Parent.(Person ∀Parent.Person ≥ 1.Parent) ≥ 1.Parent Positive (resp. negative) instances for the target concept: I + = {Meg, Gwen} (I − = {Ann, Pat}). Generated examples: P = {p 1 , p 2 } and N = {n 1 , n 2 } where
doi:10.1007/11779568_68 fatcat:5nrqg7lcezdh7lyqru7fudayoe