You are on page 1of 24

Web ontology representation and reasoning

via fragments of set theory


Domenico Cantone
Marianna Nicolosi-Asmundo

Cristiano Longo
Daniele Francesco Santamaria

Dipartimento di Matematica ed Informatica


Universit`a di Catania - ITALY
{cantone, longo, nicolosi}@dmi.unict.it, daniele.f.santamaria@gmail.com

CILC 2015 - Genova, 3 July 2015

This paper will appear in:Web Reasoning and Rule Systems, 978-3-319-22001-7, LNCS Volume 9209, Chapter
6, pp. 1-16, Springer International Publishing, 2015.

Overview of the talk


Motivation
DL4D: a description logic representable in 4LQS R
Expressiveness of DL4D
Translating SWRL rules in 4LQS R
Conclusions
Hints of future work

Motivation
Computable set theory studies the decision problem for fragments
of set theory (Cantone, Omodeo, Policriti 2001):
Integration in tnaNova/Referee;
theoretic results.
Most of the decidability results and applications concern
one-sorted multi-level fragments of set theory.
One-sorted multi-level syllogistics knowledge representation:
Cantone, Longo, Pisasale, 2010;
Cantone, Longo, Nicolosi-Asmundo, 2011;
Cantone, Longo, 2014;
... .
3

Multi-sorted stratified syllogistics


Multi-sorted stratified syllogistics are set-theoretic languages
admitting variables of different sorts (sort 0, sort 1,.. and so on)
Assignments for such variables are based on collections of objects
(natural numbers, real numbers, possible worlds...)
Variables of sort 0 objects of the considered domain
Variables of sort 1 sets of such objects
Variables of sort 2 collections of sets
...

Multi-sorted stratified syllogistics


The study of set-theoretic languages admitting variables of
different sorts is motivated by the fact that most of the formulae
in the statements and proofs of theorems in many fields of
mathematics and computer science involve variables of different
sorts
Example:
In description logics there are entities of different types: individual
elements, concepts, roles

Multi-sorted stratified syllogistics


Less investigated than one-sorted multi-level syllogistics
Some results:
Two-level syllogistics, 2LS
(A. Ferro and E. Omodeo, 1978)
Extension of 2LS with singleton and cartesian product
(D. Cantone, V. Cutello, 1990)
Three-level syllogistics, extended with powerset, general union, singleton
(D. Cantone, V. Cutello 1993)
Recently, decidability of the satisfiability problem for a four-level stratified
syllogistic, 4LQS R, admitting a restricted form of quantification over variables
of three types was proved in (D. Cantone, M. Nicolosi-Asmundo 2011)
6

This work
The description logic DL4D supporting:
datatypes,
concept constructs (i.e., concept domain and range),
role constructs (i.e., union, complement, role chains).
Decidability of the consistency problem for DL4D-knowledge
bases via a reduction to the satisfiability problem for
4LQS R-formulae.
NP-completeness of the consistency problem for h-restricted
DL4D-knowledge bases.
Expressivity of DL4D.
Representation of SWRL rules in 4LQS R.
7

Syntax of 4LQS
Pairing operator h, i
Predicate symbols = and
(i) variables
(ii) variables
(iii) variables
(iv) variables

of
of
of
of

sort
sort
sort
sort

0:
1:
2:
3:

x, y, z, . . .
X 1, Y 1, Z 1, . . .
X 2, Y 2, Z 2, . . .
X 3, Y 3, Z 3, . . .

4LQS quantifier-free atomic formulae


level 0: x = y, x X 1
level 1: X 1 = Y 1, X 1 X 2
level 2: X 2 = Y 2, hx, yi = X 2, hx, yi X 3, X 2 X 3
8

Syntax of 4LQS
4LQS quantified atomic formulae
level 1: (z1) . . . (zn)0, 0 propositional combination of
quantifier-free atomic formulae
1 ) , propositional combination of
level 2: (Z11) . . . (Zm
1 1
quantifier-free atomic formulae and of quantified atomic
formulae of level 1
level 3: (Z12) . . . (Zp2)2, 2 propositional combination of
quantifier-free atomic formulae and of quantified atomic
formulae of levels 1 and 2
4LQS-Formulae
Propositional combinations of quantifier-free atomic formulae of
levels 0, 1, 2, and of quantified atomic formulae of levels 1, 2, 3
9

Semantics of 4LQS
A 4LQS-interpretation is a pair M = (D, M ), where
Mx D
M X 1 pow(D)
M X 2 pow(pow(D))
M X 3 pow(pow(pow(D)))
We put M hx, yi = {{M x}, {M x, M y}}

10

Semantics of 4LQS
Formulae are interpreted as usual. In particular
M |= (z1) . . . (zn)0 iff M[z1/u1, . . . , zn/un] |= 0, for
all u1, . . . , un D
1 ) iff M[Z 1/U 1, . . . , Z 1 /U 1 ] |= ,
M |= (Z11) . . . (Zm
1
1
m m
1 1
1 pow(D)
for all U11, . . . , Um

M |= (Z12) . . . (Zp2)2 iff M[Z12/U12, . . . , Zp2/Up2] |= 2,


for all U12, . . . , Up2 pow(pow(D))

11

Characterizing 4LQS R
4LQS R is the subcollection of the formulae of 4LQS such that
Restriction I
Nestings of quantifiers over variables of sort 0 into quantifiers over
variables of sort 1 are allowed if the former are linked to the
corresponding variables of sort 1
1]
[ 0 ni=1 m
z

Z
i
j=1
j
Example
(Z 1)(Z 1 X 2 (z) ( z Z 1 z X 1) )
If M |= ( z Z 1 z X 1) then M |= z Z 1
12

Characterizing 4LQS R
Restriction II
Every quantified atomic formula of level 3 is
either of type (Z12), . . . , (Zp2)2, where 2 is a propositional
combination of quantifier-free atomic formulae
or of type (Z 2)(Z 2 X 3 (z1)(z2)(hz1, z2i = Z 2)
Examples
(Z 2)(Z 2 X 3 (Z 2 X13 Z 2 X23)) (intersection)
(Z 2)(Z 2 X 3 (Z 2 X13 (Z 2 X23))) (set difference)
13

The description logic DL4D


D = (ND , NC , NF , D), datatype map,
RA, RD, C, I, sets of abstract role names, concrete role
names, concept names, individual names.
(a) t1, t2 dr | t1 | t1 u t2 | t1 t t2 | {ed} ,
(b) C1, C2 A | > | | C1 | C1 t C2 | C1 u
C2 | {a} | R.Self |R.{a}|P.{ed} ,
(c) R1, R2 S | U | R1 | R1 | R1 t R2 | R1 u
R2 | RC1| | R|C1 | RC1 | C2 | id(C) ,
(d) P T | P | PC1| | P|t1 | PC1|t1 ,
14

The description logic DL4D


A DL4D-TBox is a set of statements of the types:
- C1 C2, C1 v C2, C1 v R1.C2, R1.C1 v C2,
nR1.C1 v C2, C1 v nR1.C2,
- t1 t2, t1 v t2, C1 v P1.t1, P1.t1 v C1, nP1.t1 v C1,
C1 v nP1.t1,

15

The description logic DL4D


A DL4D-RBox is a collection of statements of type:
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), P1 P2,
P1 v P2, Fun(P1).
A DL4D-ABox is a set of assertions of the forms:
a : C1, (a, b) : R1, (a, b) : R1, a = b, a 6= b, ed : t1,
(a, ed) : P1, (a, ed) : P1.

16

Semantics of terms and axioms specific to DL4D


Name
data range dr
negative datatype term
datatype terms
intersection
datatype terms union
constant in NC (d)
valued exist.
quantification
datatyped exist.
quantif.
abstract role
complement
abstract role union
abstract role
intersection
abstract role domain
restr.
concrete role domain
restr.
concrete role range
restr.
concrete role
restriction
datatype terms
equivalence
datatype terms diseq.
datatype terms
subsum.

Syntax
dr
t1

Semantics
drD D
(t1 )D = D \ tD
1

t1 u t2

D
(t1 u t2 )D = tD
1 t2

t1 t t2
ed

D
(t1 t t2 )D = tD
1 t2
D
D
ed d

R.a

(R.a)I = {x I : hx, aI i RI }

T.ed

I
(T.ed )I = {x I : hx, eD
d i T }

(R)I = (I I ) \ RI

R1 t R2

(R1 t R2 )I = R1I R2I

R1 u R2

(R1 u R2 )I = R1I R2I

RC|

(RC| )I = {hx, yi RI : x C I }

TC|

(TC| )I = {hx, yi T I : x C I }

T|t

(T|t )I = {hx, yi T I : y tD }

TC1 |t

(TC1 |t )I = {hx, yi T I : x C1I y tD }

t1 t2

D
I |=D t1 t2 tD
1 = t2

t1 6 t2

D
I |=D t1 6 t2 tD
1 6= t2

t1 v t2

D
I |=D (t1 v t2 ) tD
1 t2

17

Expressivity of DL4D
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.
Derived datatypes (inside inclusion axioms involving concrete
roles).
18

Decidability of the consistency problem for DL4D-knowledge bases


Theorem 1. Let K be a DL4D-knowledge base. Then, one can
construct a 4LQS R-formula K s.t. K is satisfiable if and only
if K is consistent.
Idea of the proof:
function from DL4D-statements into 4LQS R-formulae
1 z = x ),
Example: (C1 {a}) =Def (z)(z XC
a
1

4LQS R-formula K expressing the consistency of K,


equivalence between the consistency problem for K and the
satisfiability problem for K .
19

NP-completeness of the consistency problem for


h-restricted DL4D-knowledge bases
DL4D-knowledge base K is h-restricted if axioms of types:
R1 . . . Rn1 v R, n2R.C1 v C2, n3P.t1 v t2, C1 v n4R.C2,
t1 v n5P.t2 occur in K only if n1, n2, n3, n4, n5 h.
Reasoning as in Theorem 1, we construct 0K such that
(i) 0K belongs to (4LQS R)h, whose satisfiability problem is
NP-complete (Cantone, Nicolosi-Asmundo 2011), and
(ii) the size of 0K is polynomially related to that of K.
Quite expressive: Ontoceramic representable in (4LQS R)3.
20

Translating SWRL-rules into 4LQS R-formulae

An SWRL-rule r has the form (x1, . . . , xn)(B H), where:


- B and H are conjunctions of atoms of the following types:
x C, y t, hx, yi R, hx, yi P, x = y, x 6= y, and
- Var (H) Var (B) = {x1, . . . , xn}.

21

Translating SWRL-rules into 4LQS R-formulae


For space reasons we do not provide here a formal translation
function. However, it is not hard to see that it could be
constructed by modifying the map introduced in the proof of the
previous theorem.
Rule
hasParent(X, Y ), hasBrother(Y, Z) :
SWRL-rule
hasUncle(X, Z).
3
(x)(y)(z)(hx,
yi

X
hy, zi
R
hasParent
4LQS -rule
3
3
XhasBrother
hx, zi XhasUncle
)

Type of Rule

22

Conclusions
DL4D: a description logic expressible in 4LQS R
Decidability of the consistency problem for DL4D-knowledge
bases
NP-Completeness.
Expressiveness of SROIQ(D)
Expressing SWRL rules in 4LQS R

23

Hints of future work


DL4D: a description logic expressible in 4LQS R
Concrete roles union and intersection.
Datatype groups
Metaroles and Metaconcepts
Implementing reasoner, available on Protege.

24

You might also like