An optimized KE-tableau-based system for

reasoning in the description logic DL4,×

D

Domenico Cantone, Marianna Nicolosi-Asmundo,

Daniele Francesco Santamaria

Deparment of Mathematics and Computer Science, University of Catania

RuleML+RR 2018 - Luxembourg, 18 September, 2018

Introduction Preliminaries Reasoning Services Conclusions References

Content

Computable set theory as (web) knowledge representation and

reasoning framework.
The set-theoretic fragment 4LQSR
DL . 4,×
D

DL4,×
D : a description logic representable in 4LQSR
DL . 4,×
D

The Higher-Order Conjunctive Query Answering Problem for

DL4,×
D .
A variant of the KE-tableau-system called KEγ -tableau.
A KEγ -tableau-based algorithm for the HOCQA problem for
DL4,×
D .
Conclusions.
Introduction Preliminaries Reasoning Services Conclusions References

Some results

MLSS×
2,m , (Cantone, Longo, Nicolosi-Asmundo 2010).
DLhMLSS×
2,m i , (Cantone, Longo, Pisasale 2010).
DLh∀π0,2 i , (Cantone, Longo, Nicolosi-Asmundo 2011).
DLh∀π0,2 i + metamodelling, (Cantone, Longo 2014).
DL4D , (Cantone, Longo, Nicolosi-Asmundo, Santamaria
2015).
DL4,×
D and the Conjunctive Query Answering problem
(Cantone, Nicolosi-Asmundo, Santamaria 2016).
DL4,×
D and the Higher-Order CQA problem (Cantone,
Nicolosi-Asmundo, Santamaria 2017).
Prototype of a DL4,×
D reasoner (Cantone, Nicolosi-Asmundo,
Santamaria 2017).
Introduction Preliminaries Reasoning Services Conclusions References

Set-theoretic fragment

Syntax of 4LQSRDL 4,×

D

The fragment 4LQSR

DL . 4,×
D

(i) variables of sort 0: x, y , z, . . .

(ii) variables of sort 1: X 1 , Y 1 , Z 1 , . . .
(iii) variables of sort 3: X 3 , Y 3 , Z 3 . . .

Predicate symbols “ = ” and “ ∈ ”

Pairing operator “h·, ·i”
4LQSR
DL -literals (atomic formulae or their negations)
4,×
D

level 0: x = y , x ∈ X 1 , hx, y i ∈ X 3 ,
¬(x = y ), ¬(x ∈ X 1 ), ¬(hx, y i ∈ X 3 ).
Introduction Preliminaries Reasoning Services Conclusions References

Set-theoretic fragment

Syntax of 4LQSRDL 4,×

D

4LQSR
DL - purely universally quantified formulae
4,×
D

level 1: (∀z1 ) . . . (∀zn )ϕ0 , ϕ0 propositional combination of

(level 0) literals;

4LQSRDL -formulae: propositional combinations of (level 0) literals

4,×
D

and of purely universally quantified formulae of level 1.

Introduction Preliminaries Reasoning Services Conclusions References

Set-theoretic fragment

Semantics
A 4LQSR DL -interpretation M = (D, M):
4,×
D

MX 0 ∈ D,
MX 1 ∈ pow(D),
MX 3 ∈ pow(pow(pow(D))),
where X i ∈ Vari , for i = 0, 1, 3, and pow(s) the powerset of s.
Pair terms are interpreted à la Kuratowski:
Mhx, y i =Def {{Mx}, {Mx, My }}.
M |= ϕ is recursively defined over the structure of ϕ:
Literals are evaluated in a standard way.
Compound formulae: standard rules of propositional logic.
u ] |= ϕ0 , for all u~ ∈ D n .
M |= (∀z1 ) . . . (∀zn )ϕ0 iff M[~z /~
If M |= ϕ, then M a 4LQSR DL -model for ϕ.
4,×
D

A 4LQSDL -formula is s.t.b satisfiable if it has a 4LQSR

R
4,×
D
DL -model. 4,×
D
R R
A 4LQSDL -formula is valid if satisfied by all 4LQSDL -
4,×
D
4,×
D

interpretations.
Introduction Preliminaries Reasoning Services Conclusions References

4,×
The DL DLD

The Description Logic DL4,×

D

Existential quantification and at-least number restriction

(resp., universal quantification and at-most number
restriction) only on the left- (resp., right-) hand side of
inclusion axioms.
More liberal than SROIQ(D) in:
construction of role inclusion axioms (roles involved not
subject to any ordering relationship),
simple roles are not needed to define role inclusion axioms and
axioms involving number restrictions,
Boolean operators on roles are admitted.
The consistency problem for DL4,×D -KBs and the problem of
4,×
Higher-Order Conjunctive Query Answering for DLD are
decidable in view of the decidability of the satisfiability
problem for 4LQSR DL -formulae.
4,×
D

HOCQA via a variant of the KE-tableau system.

Introduction Preliminaries Reasoning Services Conclusions References

4,×
The DL DLD

RA , RD , C, and Ind be denumerable pairwise disjoint sets of

abstract role names, concrete role names, concept names, and
individual names, respectively. A DL4,×
D -RBox, -TBox, and -ABox
are respectively:
R1 ≡ R2 , R1 v R2 , R1 . . . Rn v Rn+1 , Sym(R1 ), Asym(R1 ),
Ref(R1 ), Irref(R1 ), Dis(R1 , R2 ), Tra(R1 ), Fun(R1 ),
R1 ≡ C1 × C2 , P1 ≡ P2 , P1 v P2 , Dis(P1 , P2 ), Fun(P1 ).
C1 ≡ C2 , C1 v C2 , C1 v ∀R.C2 , ∃R.C1 v C2 , ≥nR.C1 v C2 ,
C1 v ≤nR.C2 ,t1 ≡ t2 , t1 v t2 , C1 v ∀P.t1 , ∃P.t1 v C1 ,
≥nP.t1 v C1 , C1 v ≤nP.t1 ,
a : C1 ,(a, b) : R1 ,a = b, a 6= b, ed : t1 , (a, ed ) : P1 .
with C1 a DL4,×
D -concept term, d a datatype, t1 a datatype term,
R1 a DLD -abstract role term, P1 a DL4,×
4,×
D -concrete role term, a, b
individual names, and ed a constant in NC (d).
Introduction Preliminaries Reasoning Services Conclusions References

4,×
The consistency problem for DLD -KBs

From DL4,×
D to 4LQSRDL 4,×
D

A suitable mapping function θ is applied to KB thus yielding

the 4LQSR DL representation φKB of KB. θ maps injectively:
4,×
D

Individuals a, constants ed ∈ NC (d), ind. variables w ∈ Vi ,

and dat. variables u ∈ Ve into sort 0 variables xa , xed , xw , xu ,
the constant concepts > and ⊥, datatype terms t, concept
terms C , conc. variables c ∈ Vc , and da.t. variables t ∈ Vd
into sort 1 variables X>1 , X⊥1 , Xt1 , XC1 , Xc1 , Xt1 respectively,
the universal relation U, abstract role terms R, concrete role
terms P, ab. role variables r ∈ Var , and concrete role variable
p ∈ Vcr into sort 3 variables XU3 , XR3 , XP3 , Xr3 , Xp3 , respectively.
Vi , Ve , Vd , Vc , Var , and Vcr are pairwise disjoint
S denumerable
infinite sets of variables disjoint from Ind, {NC (d) : d ∈ ND }, C,
RA , and RD , respectively (introduced to map query variables).
Introduction Preliminaries Reasoning Services Conclusions References

4,×
The consistency problem for DLD -KBs

Let φKB be a conjunction of 4LQSR DL -literals and of 4,×

D
R
4LQSDL -formulae of type (∀z1 ) . . . (∀zn )ϕ0 with ϕ0 clause,
4,×
D

where universal quantifiers are moved as inward as possible,

and quantified variables are renamed as to be pairwise
distinct).

Example:

KB = {¬LocatedIn(Italy , Rome), Ref (isPartOf ),

LocatedIn v PartOf }

3
φKB = ¬(hxItaly , xRome i ∈ XlocatedIn )∧
3
(∀z1 )(hz1 , z1 i ∈ XisPartOf )
3 3
∧(∀z1 )(∀z2 )(¬(hz1 , z2 i ∈ XlocatedIn ) ∨ hz1 , z2 i ∈ XisPartOf )
Introduction Preliminaries Reasoning Services Conclusions References

4,×
The consistency problem for DLD -KBs

γ
KE -tableau

KEγ -tableau is a variant of KE-tableau system. The

KE-elimination and γ-rule are incorporated in a rule called
Eγ -rule and applied to universally quantified formulae.
A branch ϑ of a KEγ -tableau T is closed if either it contains
both A and ¬A, for some formula A, or a literal of type
¬(x = x). Otherwise, the branch is open.
A KEγ -tableau is closed if all its branches are closed.
Introduction Preliminaries Reasoning Services Conclusions References

4,×
The consistency problem for DLD -KBs

A formula ψ = (∀x1 ) . . . (∀xm )(β1 ∨ . . . ∨ βn ) is fulfilled in a

branch ϑ, if ϑ contains βi τ for some i = 1, . . . , n and for all τ .
A branch ϑ is fulfilled if every formula
ψ = (∀x1 ) . . . (∀xm )(β1 ∨ . . . ∨ βn ) occurring in ϑ is fulfilled.
A KEγ -tableau is fulfilled if all its branches are fulfilled.
A branch ϑ is complete if either it is closed or it is open,
fulfilled, and it does not contain any literal of type x = y ,
with x and y distinct variables.
A KEγ -tableau is complete (resp., fulfilled) if all its branches
are complete (resp., fulfilled or closed).
TKB is expanded by systematically applying the Eγ -rule and
the PB-Rule to purely universally quantified formulae until
they are fulfilled.
Introduction Preliminaries Reasoning Services Conclusions References

4,×
The consistency problem for DLD -KBs

Expansion rules

ψ S βi τ γ
E -rule PB-rule
βi τ A | A
where where A is a literal
ψ = (∀x1 ) . . . (∀xm )(β1 ∨ . . . ∨ βn ),
τ := {x1 /xo1 . . . xm /xom },
and S β i τ = Def {β 1 τ, ..., β n τ } \
{β i τ }, for i = 1, ..., n
τ is a substitution from QVar0 (ψ) = {x1 , . . . , xm } (the set of
quantified variables occurring in ψ) to Var0 (ϑ) (the set of variables
of sort 0 occurring free in ϑ).
Introduction Preliminaries Reasoning Services Conclusions References

4,×
The consistency problem for DLD -KBs

Example
Introduction Preliminaries Reasoning Services Conclusions References

4,×
The HOCQA problem for DLD

Higher-Order Conjunctive Queries

Querying the KB: Higher-Order Conjunctive Queries.

An HO DL4,× D -conjunctive query admits variables of three
sorts: individual and data type variables, concept variables,
and role variables.
An HO-DL4,×
D -conjunctive query is a conjunction of
4,×
HO-DLD -literals.
R(w1 , w2 ), P(w1 , u), C (w1 ), t(u), r(w1 , w2 ), p(w1 , u), c(w1 ),
t(u), w1 = w2 .
and their negations.
S
where w1 , w2 ∈ Vi ∪ Ind, u ∈ Ve ∪ {NC (d) : d ∈ ND }, R is
a DL4,× 4,×
D -abstract role term, P is a DLD -concrete role term,
4,× 4,×
C is a DLD -concept term, t is a DLD -datatype term,
r ∈ Var , p ∈ Vcr , c ∈ Vc , t ∈ Vd .
Introduction Preliminaries Reasoning Services Conclusions References

4,×
The HOCQA problem for DLD

Substitution
σ =Def {v1 /o1 , . . . , vn /on , e1 /ed1 , . . . , eg /edg , t1 /t1 , . . . , tl /tl ,
c1 /C1 , . . . , cm /Cm , r1 /R1 , . . . , rk /Rk , p1 /P1 , . . . , ph /Ph }
A substitution σ involving exactly the variables occurring in Q
4,×
is a solution for Q w.r.t. KB, if there exists a DLD -
interpretation I such that I |=D KB and I |=D Qσ.
The collection Σ of the solutions for Q w.r.t. KB is the
Higher-Order Answer Set of Q w.r.t. KB. Then the
higher-order conjunctive query answering problem for Q w.r.t.
KB consists in finding the HO answer set Σ of Q w.r.t. KB.
Let φ be a 4LQSR DL -formula and let ψ be a conjunction of
4,×
D
R
4LQSDL -literals. The HOCQA problem for ψ w.r.t. φ
4,×
D

consists in computing the HO answer set of ψ w.r.t. φ,

namely the collection Σ0 of all the substitutions σ 0 such that
M |= φ ∧ ψσ 0 , for some 4LQSR DL -interpretation M.
4,×
D
Introduction Preliminaries Reasoning Services Conclusions References

4,×
The HOCQA problem for DLD

ϑ is extended to DL4,×
D -HO literals:

θ(R(w1 , w2 )) =Def hxw1 , xw2 i ∈ XR3 ,

θ(P(w1 , u)) =Def hxw1 , xu i ∈ XP3 ,
θ(C (w1 )) =Def xw1 ∈ XC1 ,
θ(t(u)) =Def xu ∈ Xt1
θ(w1 = w2 ) =Def xw1 = xw2 ,
θ(c(w1 )) =Def xw1 ∈ Xc1 ,
θ(r(w1 , w2 )) =Def hxw1 , xw2 i ∈ Xr3 ,
θ(p(w1 , u)) =Def hxw1 , xu i ∈ Xp3 ,
θ(t(u)) =Def xu ∈ Xt1 ,

to HO-conjunctive queries and to substitutions in a standard way.

Introduction Preliminaries Reasoning Services Conclusions References

4,×
The HOCQA problem for DLD

Example
Introduction Preliminaries Reasoning Services Conclusions References

4,×
The HOCQA problem for DLD

Benchmarking

n
ΦKB =Def xa ∈ XD1 , xb ∈ XD1 , xc ∈ XD1 , xd ∈ XD1 , (∀z)(∀z1 )

o
(z ∈ XA1 ∧ hz, z1 i ∈ XP3 ∧ z1 ∈ XB1 ∧ hz, z1 i ∈ XP31 ) → z1 ∈ XC1 .

ΦKB generates more than 106

Introduction Preliminaries Reasoning Services Conclusions References

Future Works

More OWL serializations.

Expressiveness.
Ontology classification.
Datatype reasoning.
Plug-in for Protégé.
Ontologies expressible in 4LQS R .
Introduction Preliminaries Reasoning Services Conclusions References

References

D. Cantone and M. Nicolosi Asmundo. On the satisfiability problem for a

4-level quantified syllogistic and some applications to modal logic.
Fundamenta Informaticae, 124(4):427–448, 2013.
D. Cantone, M. Nicolosi-Asmundo, and D. F. Santamaria. A set-theoretic
approach to ABox reasoning services. In Costantini S., Franconi E., Van
Woensel W., Kontchakov R., Sadri F., Roman D. Rules and Reasoning.
RuleML+RR 2017, Lecture Notes in Computer Science, vol 10364.
Springer, 2017.
D. Cantone and M. Nicolosi-Asmundo and D. F. Santamaria. A C++
×
reasoner for the description logic DL4,
D . Proceedings of CILC 2017,
26-29 September 2017, Naples, Italy. CEUR WS, ISSN 1613-0073, Vol.
1949, pp. 276-280.
D. Cantone, M. Nicolosi-Asmundo, and D. F. Santamaria. A set-based
×
reasoner for the description logic DL4,
D . To App. in Proceedings of
SETS 2018, Southampton, United Kingdom, 5 June 2018.
Introduction Preliminaries Reasoning Services Conclusions References

Thank you