Re: Official OWL version outdated

On 5/8/2013 10:44, Dan Brickley wrote:
>
>
> On Wednesday, May 8, 2013, Holger Knublauch wrote:
>
>     Looking at the OWL version of schema.org <http://schema.org> at
>
>     http://schema.org/docs/schemaorg.owl
>
>     I notice that this seems to be a rather old version, while the
>     RDFa version
>
>     http://schema.org/docs/schema_org_rdfa.html
>
>     seems to be more recent. (When) will the OWL version be fixed?
>
>
> Is it useful? what do you prefer? The use of OWL is pretty weak since 
> we're so flexible.

It's not very useful in its current form, yet I believe it can be made 
very useful with a few changes. You guys are probably wasting an 
opportunity to get more "semantic web" people on board. My guess is that 
most OWL people look at both prominent online versions (the official one 
and the one of rdfs.org) and walk away because they are rather unusable.

Specifically, I would do the following transformations (and as an 
exercise I have actually implemented the required SPARQL updates based 
on the current OWL file):

- Clean up the owl:unionOfs with one member
- Convert any usage of schema.org datatypes with xsd ones
- Convert rdfs:range (Number or String) to xsd:float

Along with a simple instance data converter, the ontology could be 
changed to
- Replace schema:Thing with owl:Thing
- Replace schema:name with rdfs:label
- Replace schema:description with rdfs:comment
- Delete schema:url (as it's basically the URI of the subject)

Manual clean up should
- Add cardinality restrictions
- Declare owl:inverseOf relationships
- Mark outdated properties (such as the plural forms) as owl:deprecated.

Could this info be made available anywhere in machine readable form? I 
am pretty sure not only the RDF/OWL mapping could use this info.

>
> Does rdf/xml vs rdfa (or json-ld etc) matter to you? What about the 
> choice of all in one big file vs per-term?

It would be good to be able to owl:import something. The RDFa version 
does some things better than the OWL version, but not everything is 
perfect: properties with multiple rdfs:domains should use owl:unionOf (I 
guess RDFa has trouble representing this?).

And of course why not have the URIs dereferencable as true linked 
data... This should be a trivial feature to add for an organization that 
large. Even if just to show that the people behind schema.org do care 
about the semantic web community.

I am tempted to create our own copy based on the distilled RDFa version 
on some topbraid.org address because I believe there is much more 
potential here. One specific use case is that many of our customers 
build their own ontologies with concepts that are reinvented all the 
time - Person, Address etc. While our tooling is generic and can work 
with any ontology, it would be better to ship our product with some 
starter ontology and I believe schema.org could become the foundation of 
this. For this starter ontology, we would define some customized forms 
and views, e.g. so that addresses show street address above postal code 
etc. We could also define some out of the box web services with typical 
queries, reports etc. Clearly there are other product ideas in this 
space that the schema.org effort could also benefit of. The more 
alignment of data the better for everyone. Even if RDFa and Microdata 
will remain the vehicles of distributing schema.org instance data, these 
web pages may be generated by a triple store.

Sorry if this is repeating some discussions that have already happened 
elsewhere... I am trying to catch up.

HTH
Holger