The DL-Lite Family and Relations

The recently introduced series of description logics under the common moniker DL-Lite has attracted attention of the description logic and semantic web communities due to the low computational complexity of inference, on the one hand, and the ability to represent conceptual modeling formalisms, on the other. The main aim of this article is to carry out a thorough and systematic investigation of inference in extensions of the original DL-Lite logics along five axes: by (i) adding the Boolean connectives and (ii) number restrictions to concept constructs, (iii) allowing role hierarchies, (iv) allowing role disjointness, symmetry, asymmetry, reflexivity, irreflexivity and transitivity constraints, and (v) adopting or dropping the unique same assumption. We analyze the combined complexity of satisfiability for the resulting logics, as well as the data complexity of instance checking and answering positive existential queries. Our approach is based on embedding DL-Lite logics in suitable fragments of the one-variable first-order logic, which provides useful insights into their properties and, in particular, computational behavior.

💡 Research Summary

The paper conducts a systematic investigation of the computational properties of a broad family of extensions to the original DL‑Lite description logics. Starting from the lightweight DL‑Lite_R, DL‑Lite_F, and DL‑Lite_A formalisms, the authors consider five orthogonal axes of augmentation: (i) the addition of full Boolean connectives (¬, ∧, ∨) to concept expressions, (ii) the inclusion of number restrictions (≥ n, ≤ n), (iii) the allowance of role hierarchies (role inclusion axioms), (iv) the introduction of various role constraints—disjointness, symmetry, asymmetry, reflexivity, irreflexivity, and transitivity—and (v) the adoption or rejection of the Unique Name Assumption (UNA). By combining these dimensions they obtain 2⁵ = 32 distinct logics, each representing a different trade‑off between expressive power and reasoning cost.

The core methodological contribution is an embedding of every such DL‑Lite variant into a suitable fragment of the one‑variable first‑order logic (FOL¹). This translation maps Boolean operators to disjunctions and conjunctions of unary predicates, number restrictions to quantified counting formulas, role hierarchies to universal propagation rules, and role constraints to specialized unary or binary conditions. Because the complexity of satisfiability, instance checking, and positive existential query answering is already known for the relevant FOL¹ fragments, the authors can immediately infer the combined and data complexities for each DL‑Lite extension.

The results are summarized in a comprehensive table. In the baseline DL‑Lite_R (no Boolean, no counting, no role constraints, UNA assumed) satisfiability is NP‑complete in combined complexity while data complexity remains in AC⁰. Adding Boolean connectives raises combined complexity to NP‑complete but does not affect data complexity. Introducing unrestricted number restrictions pushes combined complexity up to PSPACE, and when combined with transitive roles it reaches EXPTIME. Role hierarchies alone keep data complexity at LOGSPACE and combined complexity at NP, but deep hierarchies can cause an EXPTIME blow‑up. The presence of role constraints such as symmetry or reflexivity does not increase data complexity beyond LOGSPACE, yet the combination of transitivity with counting leads to the highest complexity class (EXPTIME). Finally, dropping the UNA adds a LOGSPACE component to data complexity because equality reasoning becomes necessary, but it does not affect the combined complexity class.

From a practical standpoint, the paper offers concrete guidance for ontology engineers. If the primary concern is fast query answering over large data sets, one should avoid both Boolean connectives and unrestricted counting, and limit role constraints to those that do not interact with transitivity. When modeling requirements demand functional roles (≤ 1) or symmetric relationships, the increase in combined complexity is modest (still NP). However, any design that mixes counting with transitive roles should be approached with caution, as it may render reasoning intractable (EXPTIME). The analysis also shows that the UNA is not essential for maintaining low data complexity; its omission merely shifts data complexity from AC⁰ to LOGSPACE, which remains feasible for modern triple stores.

The paper concludes by outlining future work: (a) developing optimized reasoning algorithms that exploit the FOL¹ embedding, (b) extending the analysis to more expressive query languages (e.g., SPARQL with OPTIONAL), and (c) building tooling that automatically determines the most efficient DL‑Lite variant for a given set of modeling constraints. Overall, the study deepens our understanding of how incremental extensions to DL‑Lite affect computational behavior, providing a valuable roadmap for both theoreticians and practitioners in the Semantic Web community.