Advice on RDFS versus OWL DL, OntModel Profiles?

[Steve Vestal <st...@adventiumlabs.com>]

I have some owl:Ontology documents from different sources.  One uses
rdfs:Class and the other owl:Class.  When I glance at the standards, it
says these are equivalent in OWL Full but not in OWL DL.  There is a
discussion about this in the section "Ontology languages and the Jena
Ontology API" in the online Jena documentation.

My testing indicates that OntModel#listClasses and
OntModel#listNamedClasses will list things typed owl:Class but not
rdfs:Class.  This appears true regardless of the OntModelSpec chosen for
a reasoner, e.g., OWL_MEM_RDFS_INF, OWL_MEM_MICRO_RULE_INF,
OWL_DL_MEM_RULE_INF, all exhibit this behavior.

I see there is are OntModel#getProfile and OntModelSpec#getProfile
methods, but a hasty search did not turn up companion setProfile methods.

Two specific questions:

Is this the expected behavior for OntModel#listClasses and
OntModel#listNamedClasses?  Or is there something I'm missing?

If so, is there some way to make a simple selection between OWL_LITE,
OWL_DL, and OWL_FULL Profiles somewhere, analogous to the way different
reasoners can be selected?

To fish a bit, I would appreciate any experience and suggestions on
variations in ontologies obtained from different sources.  How many
owl:Ontology documents (ABox and TBox) turn out have a fair degree of
variability in which specific forms of assertions they contain?  How
robust or sensitive are or can applications be from a user perspective
w.r.t. variations and mixtures of profiles and choice of reasoner?

(If you are curious, this came up when looking at the OSLC standard
vocabularies.  As near as I can tell, Protege will show the classes but
restricts itself to displaying object and data properties.)

Re: Advice on RDFS versus OWL DL, OntModel Profiles?

[Luis Enrique Ramos García <lu...@googlemail.com.INVALID>]

Dear Steve,

from the questions you posted, I am going to try to answer this part:




*If so, is there some way to make a simple selection between
OWL_LITE,OWL_DL, and OWL_FULL Profiles somewhere, analogous to the way
differentreasoners can be selected?*

As i understood from documentation of jena [1],  they have partial
implementation of OWL in the reasoners they provide, that means you have to
check carefully the provided table to determine if the constructors you
require are implemented.
and the recommend to use an external reasoner for owl dl.

All available reasoners are in the reference, unless there is a more recent
information.


Bets regards


Luis Ramos

[1] https://jena.apache.org/documentation/inference/







El dom., 19 abr. 2020 a las 16:16, Steve Vestal (<
steve.vestal@adventiumlabs.com>) escribió:

> I have some owl:Ontology documents from different sources.  One uses
> rdfs:Class and the other owl:Class.  When I glance at the standards, it
> says these are equivalent in OWL Full but not in OWL DL.  There is a
> discussion about this in the section "Ontology languages and the Jena
> Ontology API" in the online Jena documentation.
>
> My testing indicates that OntModel#listClasses and
> OntModel#listNamedClasses will list things typed owl:Class but not
> rdfs:Class.  This appears true regardless of the OntModelSpec chosen for
> a reasoner, e.g., OWL_MEM_RDFS_INF, OWL_MEM_MICRO_RULE_INF,
> OWL_DL_MEM_RULE_INF, all exhibit this behavior.
>
> I see there is are OntModel#getProfile and OntModelSpec#getProfile
> methods, but a hasty search did not turn up companion setProfile methods.
>
> Two specific questions:
>
> Is this the expected behavior for OntModel#listClasses and
> OntModel#listNamedClasses?  Or is there something I'm missing?
>
> If so, is there some way to make a simple selection between OWL_LITE,
> OWL_DL, and OWL_FULL Profiles somewhere, analogous to the way different
> reasoners can be selected?
>
> To fish a bit, I would appreciate any experience and suggestions on
> variations in ontologies obtained from different sources.  How many
> owl:Ontology documents (ABox and TBox) turn out have a fair degree of
> variability in which specific forms of assertions they contain?  How
> robust or sensitive are or can applications be from a user perspective
> w.r.t. variations and mixtures of profiles and choice of reasoner?
>
> (If you are curious, this came up when looking at the OSLC standard
> vocabularies.  As near as I can tell, Protege will show the classes but
> restricts itself to displaying object and data properties.)
>
>
>