Re: Does rdfs:superClassOf (or equivalent) exist?

I'm afraid you can't have it both ways, both forms are semantically
"about" odrl:Policy in the same way, regardless of whether or not you use
an inverse property.

(for "aboutness" I am going with the OWLAPI's definition, but YMMV)

There is an interesting sociotechnical aspect to this, regarding the norms
for making assertions about other people's resources. The fact I can assert
whatever superclasses I like about schema:Person in my own graphs is a
feature of RDF, but individual communities may have norms and guidelines
about best practices for doing this in a way that avoids unhappiness and
chaos. This has been a challenge for OBO, where originally it was the norm
for anyone to say anything about anything, but we are slowly converging on
a "no semantic injections" policy:
https://github.com/OBOFoundry/OBOFoundry.github.io/issues/1443

Of course, syntactical or structural sleights of hand in flipping the
asserted direction of axioms don't help if the entailment remains the same
:-)

On Mon, Oct 30, 2023 at 1:24 PM Harshvardhan J. Pandit <me@harshp.com>
wrote:

> Hi All.
> (as always) Thank you for your thoughts and collective experience(s).
> Perhaps I should explain my question and thinking better here which is
> that superClassOf has its uses.
>
> My use-case is creating taxonomies where I want to express relations
> with external vocabularies/ontologies. For example, I have dpv:Policy in
> the DPV ontology and I want to express that odrl:Policy from the ODRL
> ontology is a subclass of this. In the RDF serialisation with
> rdfs:subClassOf the triples related to dpv:Policy are not in a single
> place because the subjects are different. E.g.
>
> dpv:Policy skos:prefLabel "Policy"@en .
> # 1000 lines later
> odrl:Policy rdfs:subClassOf dpv:Policy .
>
> Instead, I find it nicer and semantically correct to have:
>
> dpv:Policy rdfs:superClassOf odrl:Policy.
>
> where no reasoning or querying is required, the information I *want* to
> express is recorded exactly, and I am also not making statements about
> IRIs not under my control. Regarding direction of properties, I like how
> SKOS gives both broader/narrower even if they are inverse relations of
> each other - because it allows us to make appropriate statements about
> the 'subject' concept instead of just using broader and switching
> subject and object.
>
> I understand if such use was not 'foreseen' in RDF/RDFS, and if it is
> considered to be too niche or trivial. However, I find it immensely
> helpful when working with the concepts and especially when I am
> producing documentation as the RDF matches the through process, e.g.
> description of a concept as all triples where it is the subject.
>
> --- further notes ---
>
> I am well aware that we can use SPARQL and other querying/reasoning to
> derive parent/super classes using rdfs:subClassOf. However, because
> there is no mechanism to *store* the results, I have to run this query
> every single time I want a concept's sub-classes. e.g. given the graph:
>
> :A rdfs:subClassOf :X .
> :B rdfs:subClassOf :X .
>
> I have to run this query every time: `SELECT ?c WHERE { ?c
> rdfs:subClassOf X }` - which means I *must* have a query engine -
> instead if there was a way to 'store' ?c (which is the superclass
> relation) - then I only need a RDF parser.
>
> This also makes a difference when considering discrete sets of
> information, e.g. in the earlier example, if A and B are in graph G1 and
> X is in graph G2, then the only way for X to express A and B are its
> subclasses is to include those triples from G1 in G2. Instead, having a
> superclass relation with X as the subject here is better aesthetics and
> semantics:
>
> G1 { # A and B are 'native' to this graph
> :A rdfs:subClassOf :X .
> :B rdfs:subClassOf :X .
> }
>
> G2 { # X is 'native' to this graph
> :A rdfs:subClassOf :X . # we have no 'control' over :A
> :B rdfs:subClassOf :X . # we have no 'control' over :B
> :X rdfs:superClassOf :A, :B . # we have 'control' over :X
> }
>
> ---
>
> Regards,
> Harsh
>
> On 30/10/2023 16:52, Miguel wrote:
> > Hi Harshvardhan,
> > just to further elaborate, with that argument you could support the
> > explicit definition of the inverse of any object property (excluding
> > symmetric properties).
> > Some vocabularies do indeed adopt that convention.
> >
> > The problem with that is that at the syntactic level you have always two
> > ways (for each pair of related resources) to represent the same meaning.
> > Only if you explicitly state in OWL the relationship between the two
> > properties with the inverseOf expression *and* perform inference, then
> > the two versions are reconciled.
> >
> > Furthermore, the direction in which a property is defined does not (at
> > least in theory) imply a specific favoured direction for traversing it,
> > as shown in the SPARQL examples by Antoine Zimmermann.
> >
> > Best regards,
> > Miguel
> >
> > On Mon, Oct 30, 2023 at 5:47 PM Antoine Zimmermann
> > <antoine.zimmermann@emse.fr <mailto:antoine.zimmermann@emse.fr>> wrote:
> >
> >     rdfs:superClassOf (respectively rdfs:superPropertyOf) does not exist
> in
> >     any standard, nor any term equivalent to it.
> >
> >     If it did, it would not add anything to reasoning or querying. If you
> >     want to list all subclasses of a class <C>, you can write:
> >
> >        SELECT ?subclass WHERE {
> >          ?subclass  rdfs:subClassOf  <C> .
> >        }
> >
> >     and if you want the list of superclasses of a class <C>, you write:
> >
> >        SELECT ?superclass WHERE {
> >          <C>  rdfs:subClassOf  ?superclass .
> >        }
> >
> >
> >     What use cases would make it harder and more painful to write:
> >
> >        :A  rdfs:subClassOf  :B
> >
> >     than:
> >
> >        :B  rdfs:superClassOf  :A
> >
> >     ?
> >
> >
> >     --AZ
> >
> >     Le 28/10/2023 à 14:21, Harshvardhan J. Pandit a écrit :
> >      > Hi.
> >      > We have rdfs:subClassOf defined in a standardised specification
> >     (RDFS).
> >      > RDFS several times mentions 'superclass', but AFAIK there is no
> >     property
> >      > or relation to make this explicit, i.e.
> >      >
> >      > ```turtle
> >      > :A rdfs:subClassOf :B . # exists
> >      > :B rdfs:superClassOf :A . # does this exist anywhere?
> >      > ```
> >      >
> >      > I can intuit why subclass relations are the most common and
> >     preferred
> >      > methods of use - because anyone can extend the superclass from
> >     anywhere.
> >      > And that either assertion can be inferred from the other (sub to
> >     super,
> >      > vice-versa), but I also think having the superclass be 'aware' of
> >      > subclasses is a good practice in maintaining ontologies e.g. to
> >     get a
> >      > list of all subclasses which would normally require a query each
> >     time.
> >      >
> >      > (Likewise for rdfs:subPropertyOf and rdfs:superPropertyOf)
> >      >
> >      > Apologies in advance if this has already been answered somewhere
> (I
> >      > would appreciate it if you point me to it).
> >      >
> >      > Regards,
> >
> >     --
> >     Antoine Zimmermann
> >     École des Mines de Saint-Étienne
> >     158 cours Fauriel
> >     CS 62362
> >     42023 Saint-Étienne Cedex 2
> >     France
> >     Tél:+33(0)4 77 49 97 02
> >     http://www.emse.fr/~zimmermann/ <http://www.emse.fr/~zimmermann/>
> >
>
> --
> ---
> Harshvardhan J. Pandit, Ph.D
> Assistant Professor
> ADAPT Centre, Dublin City University
> https://harshp.com/
>
>