next up previous
Next: Expressing Cardinality Restrictions Using Up: The Complexity of Reasoning Previous: Reducing Domino Problems to

   
Reasoning with Nominals

Nominals, i.e., atomic concepts referring to single individuals of the domain, are studied both in the area of DLs [BPS94,DLNS96] and modal logics [GG93,BS96,ABM99]. In this section we show how, in the presence of nominals, consistency for \ensuremath {\text {T\hspace {-1pt}}_C\hspace {-1pt}\text {Boxes}} can be polynomially reduced to consistency of TBoxes consisting of general inclusion axioms, which, in general, is an easier problem. This correspondence is used to obtain two novel results: (i) we show that, for unary coding, consistency of \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}}-TBoxes consisting of cardinality restrictions can be decided in \textsc{Exp\-Time}; (ii) we show that, in the presence of both inverse roles and number restrictions, reasoning with nominals is strictly harder than without nominals: the complexity of determining consistency of TBoxes with general axioms rises from \textsc{Exp\-Time} to \textsc{NExp\-Time}, and the complexity of concept satisfiability rises from \textsc{PSpace} to \textsc{NExp\-Time}.

Definition 4.1   Let NI be a set of individual names (also called nominals) disjoint from NC and NR. Concepts in \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}\!\mathcal{O}} are defined as \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}}-concepts with the additional rule that, for every $o \in N_I$, o is an \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}\!\mathcal{O}}-concept. A general concept inclusion axiom for \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}\!\mathcal{O}} is an expression of the from $C \sqsubseteq D$, where C and D are \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}\!\mathcal{O}}-concepts. A \ensuremath{\text{T\hspace{-1pt}}_I\hspace{-1pt}\text{Box}} for \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}\!\mathcal{O}} is a finite set of general inclusion axioms for \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}\!\mathcal{O}}, where the subscript ``I'' stands for ``Inclusion''. The semantics of \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}\!\mathcal{O}} concepts is defined similar as for \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}}, with the additional requirement that every interpretation maps a nominal $o \in N_I$ to a singleton set $ o^\mathcal{I}\subseteq \Delta^\mathcal{I}$; o can be seen as a name for the element in $o^\mathcal{I}$. Please note that we do not have a unique name assumption, i.e., we do not assume that $o_1 \neq o_2$ implies $o_1^\mathcal{I}\neq o_2^\mathcal{I}$.

An interpretation $\mathcal{I}$ satisfies an axiom $C \sqsubseteq D$ iff $C^\mathcal{I} \subseteq D^\mathcal{I}$. It satisfies a \ensuremath{\text{T\hspace{-1pt}}_I\hspace{-1pt}\text{Box}} $\ensuremath{T_{\text{gci}}}\xspace$ iff it satisfies all axioms in $\ensuremath{T_{\text{gci}}}\xspace$; in this case $\mathcal{I}$ is called a model of $\ensuremath{T_{\text{gci}}}\xspace$, and we will denote this fact by $\mathcal{I}\models \ensuremath{T_{\text{gci}}}\xspace$. A \ensuremath{\text{T\hspace{-1pt}}_I\hspace{-1pt}\text{Box}} that has a model is called consistent.

Cardinality restrictions, \ensuremath {\text {T\hspace {-1pt}}_C\hspace {-1pt}\text {Boxes}}, and their interpretation for \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}\!\mathcal{O}} are defined analogously to \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}}.

With $\ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{O}}\xspace$ we denote the fragment of $\ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}\!\mathcal{O}}\xspace$ that does not contain any inverse roles R-1.

Lemma 4.2
Consistency of \ensuremath {\text {T\hspace {-1pt}}_C\hspace {-1pt}\text {Boxes}} or \ensuremath{\text{T\hspace{-1pt}}_I\hspace{-1pt}\text{Boxes}} both for \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{O}} and \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}\!\mathcal{Q}\!\mathcal{I}\!\mathcal{O}} is \textsc{Exp\-Time}-hard and can be decided in \textsc{NExp\-Time}, if unary coding of numbers is used.

Proof. Consistency of \ensuremath{\text{T\hspace{-1pt}}_I\hspace{-1pt}\text{Boxes}} (and hence of \ensuremath {\text {T\hspace {-1pt}}_C\hspace {-1pt}\text {Boxes}}) is \textsc{Exp\-Time}-hard already for (a syntactical variant of) \ensuremath{\mathcal{A}\!\mathcal{L}\!\mathcal{C}} [HM92]. Assuming unary coding of numbers, we can reduce the problems to satisfiability of \ensuremath{C^2}, which yields the \textsc{NExp\-Time} upper bound.



 
next up previous
Next: Expressing Cardinality Restrictions Using Up: The Complexity of Reasoning Previous: Reducing Domino Problems to

Stephan Tobies
May 02 2000