Re: Official OWL version outdated

Hi Holger,

On 5/16/2013 12:10 AM, Holger Knublauch wrote:
> Hi Bo,
>
> thanks for your feedback.
>
> On 5/16/2013 4:56, Bo Ferri wrote:
>> Hi Holger,
>>
>> thanks a lot for providing this well-thoughtout schema.org to proper
>> RDF(S) mapping. Good work! I really endorse the mappings for the
>> schema/ontological terms that make use of the well known terms from
>> OWL and RDFS (instead of schema:Class etc.).
>> However, I'm wondering why you've introduced a new name space
>> 'schemax' to create new classes, whose ranges could also be modelled
>> via owl:unionOf, or?
>
> Yes I agree this isn't perfect and that's also the main reason why I
> marked the current version as "experimental". I would like to get more
> feedback on this.
>
> The main reason for just having a single range with named union classes
> is that many tools (including some of our own) wouldn't know what type
> of widget to display and how to best allow users to select values.

Yes, okay. You made the point. So one has to decide between defacto best 
practice ontology design and "real world" practical applicability. I'm 
not sure, however, I think since popular ontologies such as 
GoodRelations are also doing it this way, there must be a practical 
solution to implement it with a satisfiable UX, or? Albeit, I guess you 
are looking for a rather generic approach here ... where it might get a 
bit more complicated ;)

> This
> problem is especially obvious for mixed properties that allow xsd:string
> or objects.
>
> I will rename the current version (with the single range) and change the
> converter to use owl:unionOf ranges, then replace the version to use
> that convention. Then people have the choice which range convention they
> use.

coolo!

>
> Question: Some properties either allow objects or xsd:string values.
> Shall I leave those in tact or shall I suppress the xsd:string choice?
> If I leave them in I would need to turn them into rdf:Properties. I have
> not seen properties that have such a mixed union anywhere, but I guess
> it would be more correct to preserve them. Another idea would be to
> introduce different properties for the string part, e.g.
> schema:xy-string, but this would introduce different URIs. Please let me
> know your thoughts.

I would endorse the rather generic, not OWL-DL compatible way by 
utilising rdf:Property for relationships that could be datatype or 
object properties. However, I know that this a decision against OWL-DL 
(see also this discussion [1]). Two properties, which bare the same 
meaning, "only" for OWL-DL compatibility will confuse the common user of 
the vocabulary. I know that one could also implement OWL property chain 
axioms or similar mechanism, but I guess that this would be a bit 
against the nature of schema.org (which is nowadays already to big to 
grasp for the common web developer (re. number of types etc.) - but this 
is another topic ... ;) ).

Cheers,


Bo


[1] 
http://answers.semanticweb.com/questions/1367/restrictions-on-rdfproperty

Received on Thursday, 16 May 2013 19:32:21 UTC