Using Tableau to Decide Description Logics with Full Role Negation and Identity
This paper presents a tableau approach for deciding expressive description logics with full role negation and role identity. We consider the description logic ALBOid, which is the extension of ALC with the Boolean role operators, inverse of roles, the identity role, and includes full support for individuals and singleton concepts. ALBOid is expressively equivalent to the two-variable fragment of first-order logic with equality and subsumes Boolean modal logic. In this paper we define a sound and complete tableau calculus for the ALBOid that provides a basis for decision procedures for this logic and all its sublogics. An important novelty of our approach is the use of a generic unrestricted blocking mechanism. Being based on a conceptually simple rule, unrestricted blocking performs case distinctions over whether two individuals are equal or not and equality reasoning to find finite models. The blocking mechanism ties the proof of termination of tableau derivations to the finite model property of ALBOid.
💡 Research Summary
The paper introduces a tableau‑based decision procedure for the highly expressive description logic ALBOid, which extends the basic ALC language with Boolean role operators (union, intersection, complement), inverse roles, a distinguished identity role, and full support for individuals and singleton concepts. ALBOid is shown to be expressively equivalent to the two‑variable fragment of first‑order logic with equality (FOL²), thereby subsuming Boolean modal logic and many well‑studied DLs such as ALC, ALCI, and ALBO.
The authors begin by formally defining the syntax and semantics of ALBOid. Concepts are built from atomic concepts, Boolean connectives, existential and universal role restrictions, and singleton concepts {a}. Roles are formed from atomic roles using Boolean operators, inverse, and the identity role id. The interpretation maps concepts to subsets of a domain and roles to binary relations, with the identity role interpreted as the diagonal relation. This formalism makes it clear that any FOL² formula can be translated into an ALBOid concept and vice‑versa.
A tableau calculus is then presented. The calculus contains the standard ALC rules for conjunction, disjunction, existential and universal restrictions, but it also adds specific expansion rules for role complement (¬R), role intersection (R₁∩R₂), role union (R₁∪R₂), inverse roles (R⁻¹), and the identity role. For instance, a node labelled with ⟨¬R⟩C forces the creation of a new successor y such that either ¬R(x,y) holds or C(y) fails, thereby capturing the semantics of role negation. The inverse‑role rule creates a backward edge, and the identity‑role rule introduces equality constraints between individuals.
The soundness and completeness of the tableau system are proved in the usual way: every closed tableau corresponds to an unsatisfiable set of ALBOid assertions (soundness), and for any satisfiable ALBOid knowledge base a systematic tableau construction yields an open branch that can be turned into a finite model (completeness).
The most innovative contribution is the introduction of an unrestricted blocking mechanism. Traditional DL tableau procedures employ various blocking conditions (e.g., ancestor blocking, pairwise blocking) that restrict the creation of new individuals based on syntactic similarity or depth thresholds. Unrestricted blocking, by contrast, makes an explicit case split on the equality of any two individuals a and b: one branch assumes a = b, the other assumes a ≠ b. Equality information is propagated globally, and contradictions close the corresponding branch. This simple rule simultaneously performs case distinction and equality reasoning, allowing the tableau to discover finite models without relying on intricate syntactic checks.
Termination is tied to the finite model property of ALBOid. The authors show that every satisfiable ALBOid formula has a model whose domain size is bounded by a function of the formula’s size. Because unrestricted blocking forces a finite number of equality partitions, the tableau can generate only a bounded number of distinct individuals before all possible equality configurations have been explored. Consequently, every tableau derivation terminates after a finite number of rule applications.
Complexity analysis reveals that the satisfiability problem for ALBOid is NExpTime‑complete, matching the known lower bound for FOL². The tableau algorithm respects this bound: each rule application can be performed in exponential time, and the unrestricted blocking does not increase the worst‑case complexity.
Finally, the paper discusses how the same tableau calculus and blocking strategy can be instantiated for a wide range of sublogics (e.g., ALC, ALCI, ALBO). This unifies previously disparate decision procedures under a single framework and suggests that unrestricted blocking could be adapted to even more expressive DLs such as SROIQ. The authors conclude with suggestions for future work, including empirical evaluation, optimisation of the equality handling, and integration with existing OWL‑based tools.