Fitting Horn DL Ontologies to ABox and Query Examples: A Tale of Simulation Quantifiers and Finite Models
Pith reviewed 2026-05-07 14:50 UTC · model grok-4.3
The pith
Simulations between examples and models decide whether a fitting Horn DL ontology exists for given ABox and query data.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
Existence of a fitting ontology is equivalent to the existence of finite models that simulate the positive examples while failing to simulate the negative ones, under the query language in use. For atomic queries this check runs in polynomial time in both EL and ELI. For rooted conjunctive queries and unions of them the same check is Sigma_P^2-complete in EL and ExpTime-complete in ELI. Extending the logics by the bottom concept leaves all listed complexities unchanged.
What carries the argument
Simulation relations (and their quantifier variants) between the given ABox examples and finite candidate models, which serve as the decision criterion once the finite-model property is invoked.
If this is right
- Atomic-query fitting is tractable and admits practical algorithms for both EL and ELI.
- Rooted conjunctive-query fitting jumps to Sigma_P^2-completeness in EL, indicating that NP oracles are likely required.
- The same problem reaches ExpTime-completeness in ELI, matching the complexity of standard reasoning tasks in that logic.
- Inclusion of the bottom concept does not raise or lower any of the established bounds.
Where Pith is reading between the lines
- Practical fitting tools could therefore be built for atomic queries without heavy computational overhead.
- The jump in complexity when moving from atomic to conjunctive queries suggests that query-language choice will be a key design parameter in any deployed system.
- The finite-model property is load-bearing, so extensions to query languages without this property would require separate analysis.
Load-bearing premise
Fitting reduces to checking whether finite simulation relations exist between the supplied examples and suitably chosen models.
What would settle it
An ABox-plus-query pair for which a fitting EL ontology exists, yet every pair of finite models that simulates the positive examples also simulates the negative ones.
read the original abstract
We study the problem of fitting a description logic (DL) ontology to a given set of positive and negative examples that take the form of an ABox and a Boolean query. While previous work has investigated this problem for the expressive DLs ALC and ALCI, we here focus on the Horn DLs EL and ELI, as well as their extensions with the bottom concept. As the query language, we consider atomic queries (AQs), conjunctive queries (rooted CQs), and unions thereof (rooted UCQs). We provide characterization of the existence of a fitting ontology based on simulations, use them to develop decision procedures, and clarify the exact computational complexity. For AQs, the problem is in PTime for both EL and ELI. For rooted CQs and UCQ, it is Sigma_P^2-complete for EL and ExpTime-complete for ELI. Adding the bottom concept does not change any of these complexities. Interestingly, moving from ALC and ALCI to EL and ELI introduces additional technical challenges rather than simplifying the matter.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. This paper examines the ontology fitting problem for Horn description logics EL and ELI, with and without the bottom concept, where examples are provided as an ABox together with a Boolean query (atomic, rooted conjunctive, or union of rooted conjunctive queries). It offers characterizations of when a fitting ontology exists using simulation relations between models, derives decision procedures from these characterizations, and determines the computational complexity of the problem for each combination of logic and query language: PTime for atomic queries in both EL and ELI; Σ₂ᵖ-complete for rooted CQs and UCQs in EL; and EXPTIME-complete for rooted CQs and UCQs in ELI. Adding the bottom concept leaves all complexities unchanged.
Significance. If the simulation characterizations and complexity results hold, the work provides a complete and precise complexity classification for the fitting problem in practically relevant Horn DLs. The simulation-based approach, grounded in first-principles relations and the finite-model property, yields decision procedures and highlights that restricting to Horn fragments introduces additional technical challenges compared to ALC/ALCI. This advances the field by supplying both theoretical characterizations and algorithmic insights that can support the development of fitting tools in knowledge representation.
minor comments (3)
- The abstract states 'rooted CQs and UCQ' but should read 'rooted CQs and UCQs' for consistent pluralization.
- The introduction's reference to prior ALC/ALCI fitting results would benefit from explicit citations to the specific papers establishing those baselines.
- Definitions of the simulation quantifiers in the characterizations would be clearer with a small illustrative example or diagram.
Simulated Author's Rebuttal
We thank the referee for the positive assessment of our work, the recognition of its significance for Horn DL ontology fitting, and the recommendation of minor revision. We appreciate the confirmation that our simulation-based characterizations and complexity results provide a complete classification for EL/ELI with ABox and query examples.
Circularity Check
No significant circularity in derivation chain
full rationale
The paper derives characterizations of the existence of fitting Horn DL ontologies (EL/ELI with/without bottom) to ABox and query examples directly from simulation relations between examples and candidate models, together with the finite-model property for these fragments. These characterizations are used to obtain decision procedures whose complexities (PTime for AQs, Sigma_P^2-complete for rooted CQs/UCQs in EL, ExpTime-complete in ELI) follow from standard complexity reductions on the simulation checks. No step reduces a claimed result to a fitted parameter, self-definition, or load-bearing self-citation by construction; the simulation-based foundation is independent of the target fitting existence and complexity statements.
Axiom & Free-Parameter Ledger
axioms (2)
- standard math Simulation relations preserve concept and role structure between models in the relevant DLs
- domain assumption Finite models suffice for the decision problems involving the given query languages
Reference graph
Works this paper leans on
-
[1]
Proof.‘1 ⇒ 2’
there exists a completionCofA − such that (a)Q(a)̸∈ Cfor each(A, Q(a))∈E −; (b) if (A, Q(a))∈E + and (A, a)⪯ L (C, b), then Q(b)∈ C. Proof.‘1 ⇒ 2’. Suppose that E admits a fitting L- ontology O. Then for every e= (A, Q(a))∈E − there is a model Ie of A and O with Ie ̸|=Q(a) . Since L- ontologies are invariant under disjoint union, the interpre- tation I= U...
-
[2]
This is only possible if L ∈ {EL ⊥,ELI ⊥} because then simulations are required to be total
There is noL-simulation fromAtoC. This is only possible if L ∈ {EL ⊥,ELI ⊥} because then simulations are required to be total. As in the ‘(ii)⇒(i)’ direction of the proof of Theorem 1, we find an aA ∈ ind(A)such that for allb∈ind(C), (A, aA)̸⪯ ℓ L (A+, b) whereℓ=|C| · |A|. We then set Oe :={C ℓ,L A,aA ⊑ ⊥}
-
[3]
We set Oe :={C ℓ,L A,a ⊑Q}, whereℓ=|C| · |A|
There is anL-simulation fromAtoC. We set Oe :={C ℓ,L A,a ⊑Q}, whereℓ=|C| · |A|. We show that O:= S e∈E+ Oe fits E. For each e= (A, Q(a))∈E +, we have A ∪ Oe |=Q(a) , thus A ∪ O |= Q(a) as required. In both cases this is due to b∈(C ℓ,L A,b)J for all modelsJofAand individualsb∈ind(A). Next consider a negative example e= (A, Q(a))∈E −. Let IC be C viewed as...
-
[4]
If I is an L-forest model of A, then I |=p if and only if there exists an L-forest A-variation p′ of p such that I |=p ′
-
[5]
Proof.We start with Point 1
for every L-forest A-variation p′ of p: I |=p ′ if and only ifI |=bp♯ for allbp♯ ∈red(p ′). Proof.We start with Point 1. ‘only if’. Let I be an L- forest model of A and suppose that I |=p , witnessed by a homomorphism g from p to I. Using g, we construct an L- forest A-variation p′ of p as follows. First, for every variable x with g(x) =a∈ind(A) , replace...
-
[6]
there is anL-ontology that fitsE
-
[7]
there exists a finite interpretationIsuch that (a)I= U e∈E− Ie where, for each e= (A, q)∈E −, Ie is a model ofAwithI e ̸|=q; (b) for all (A, q)∈E +: if S is an L-simulation from A to I, thenI A,S,L |=q. To prove the theorem, we need a way to handle the non- rooted parts of the queries in Condition (b) uniformly when constructing a fitting ontology from a ...
2018
-
[8]
Observe that then L=L ′ ⊥ ∈ {EL ⊥,ELI ⊥} and there exists an indi- vidual a∈ind(A) such that (A, a)̸⪯ L′ (I, d), for all d∈∆ I
There exists no L-simulation from A to I. Observe that then L=L ′ ⊥ ∈ {EL ⊥,ELI ⊥} and there exists an indi- vidual a∈ind(A) such that (A, a)̸⪯ L′ (I, d), for all d∈∆ I. To guarantee that the constructed ontology fits such an example, we must ensure that A is inconsistent withO. Hence, we set ens(e) :={⊥(a)}
-
[9]
By assumption and Lemma 10, there exists some rooted CQ p in q and an L-forest A-variation p′ of p such that IA,S,L |=p ′
Otherwise, let S be the maximal L-simulation from A to I. By assumption and Lemma 10, there exists some rooted CQ p in q and an L-forest A-variation p′ of p such that IA,S,L |=p ′. Since S is chosen to be maxi- mal, it suffices to ensure that A ∪ O entails the reduc- tion queries of p′. For each existential reduction query bp♯ =∃x C(x)∈red(p ′), let abp♯ ...
-
[10]
there exists a finite interpretation I that satisfies Condi- tions (a′) and (b′) of Lemma 7
-
[11]
memorized
for every e∈E + there exists an Oe ∈Γ e such that A ∪S e∈E− Oe ̸|=fin p′ for every (A, q)∈E − and L- forestA-variationp ′ of a CQ inq. ‘1⇒ 2’ Assume there exists finite modelI that satisfies Conditions (a′) and (b′) of Lemma 7. Since I satisfies Con- dition (b′), for every e∈E + there exists an Oe ∈Γ e such that I |=S e∈E+ Oe. Let e= (A, q)∈E −, and Ie be...
2004
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.