Re: [OWLAPI-developer] Strange behavior: entailment of non-OWL 2 DL property characteristics

[Despoina M. <des...@li...>]

Hello,

As Ian pointed out in previous discussions,
this is an inherent problem when adding axioms to explicate implicit
characteristics of properties. A similar situation appears in the following 
example. Assume the axiom:

Thing subClassOf (allValuesFrom R (allValuesFrom R Nothing))

which asserts for some role R that there don't exist any role chains
longer than one. From this we can infer that R is transitive. However,
it is still fine to use R in cardinality constraints -- this won't
push the ontology out of OWL 2 DL. Adding an axiom that explicitly
asserts the transitivity will, however, push the axiom out of OWL DL.

This "problem" arises because the syntactic conditions used to define
simple roles (and regular hierarchies) are sufficient but not
necessary for ensuring decidability. The "solution" is, as Birte said,
not to explicate implicit characteristics of properties that would
make the property hierarchy irregular, that is, not to add the inferred 
axioms to the input ontology.

Best,
Despoina

--
Despoina Magka, Room 309
Research student
Oxford University Computing Laboratory
OX1 3QD
United Kingdom
+44 (0)1865 283520
________________________________________
From: Rinke Hoekstra [rjh...@gm...] On Behalf Of Rinke Hoekstra [hoe...@uv...]
Sent: 24 May 2011 09:09
To: Pellet; su...@he...; owl...@li...
Cc: Jochem Douw
Subject: Strange behavior: entailment of non-OWL 2 DL property characteristics

(apologies for the crosspost, but this issue occurs both with Pellet 2.2.2 and Hermit 1.3.4 using the OWL API)

Hi all,

I have an issue with the way in which the OWL API reports the property characteristics entailed by Pellet and Hermit. Sort form: in some cases both reasoners report the entailment of property characteristics that push the ontology outside OWL 2 DL.

Suppose I have the following ontology O that defines a property 'isTransitiveTest' with disjoint domain and range:

Declaration(Class(:A))
Declaration(Class(:B))
DisjointClasses(:A :B)
DisjointClasses(:B :A)
Declaration(ObjectProperty(:isTransitiveTest))
ObjectPropertyDomain(:isTransitiveTest :A)
ObjectPropertyRange(:isTransitiveTest :B)


Obviously, this ontology is consistent and coherent.

Both Pellet and Hermit will report a positive entailment for the following property characteristics:

InverseObjectProperties(:isTransitiveTest ObjectInverseOf(:isTransitiveTest))
AsymmetricObjectProperty(:isTransitiveTest)
TransitiveObjectProperty(:isTransitiveTest)
IrreflexiveObjectProperty(:isTransitiveTest)


This ontology is no longer in OWL 2 DL as the non-simple property (i.e. the transitive property :isTransitiveTest) occurs in an asymmetric object property axiom.

Note that the reasoners do not report any such problems with ontology O, which means (I guess) that the property characteristics are not used in reasoning unless explicitly asserted. So, somehow the entailment check used by the OWL API produces more entailments than KB realization.

Cheers,
Rinke


---
Dr Rinke Hoekstra

AI Department (Guest) |  Leibniz Center for Law
Faculty of Sciences  |  Faculty of Law
Vrije Universiteit |  Universiteit van Amsterdam
De Boelelaan 1081a |  Kloveniersburgwal 48
1081 HV Amsterdam  |  1012 CX Amsterdam
+31-(0)20-5987752  |  +31-(0)20-5253497
r.j...@vu... |  hoe...@uv...

Homepage: http://www.few.vu.nl/~hoekstra
Sent with Sparrow