Re: Returning to OWL and application profiles

Hi Tom,


> On Tue, Oct 12, 2010 at 05:09:12PM +0200, Emmanuelle Bermes wrote:
>>                         We are lacking actual use of Linked Data today,
>> and I feel that adding more complexity in the data model is likely to
>> create more barriers.
>
> There's the point.  Is it really necessary to create more
> classes, more properties, and more triples in the data that is
> actually published?


Yes. Please believe that I'd feel very concerned about proliferation of classes or properties, too. If that would be the price to pay then indeed this is not a very seducing solution. But I don't think anyone is suggesting that all properties should be duplicated in APs. As a matter of fact I believe that many examples in this thread could be dealt with by just introducing one AP-specific class, and putting axioms for it that just re-use the existing properties. Which is of uttermost importance: on the linked data environment being able to re-use the link types is much more important than re-using the classes (which is an aspect which I believe also emerged in our previous discussion on FRBR).

And of course again there is the possibility of trying to implement constraints via SPARQL. As long as we can relate to an established bit of the SW architecture that would be already quite positive, I think.
  
  
>> I like the idea that the AP should be something that could be
>> implemented following different syntaxes, maybe including OWL, but not
>> excluding other approaches that wouldn't make it mandatory to declare
>> local properties and classes systematically when additional semantics
>> or constraints are needed.
>
> I also wonder about the training angle.  I think we want to
> enable people to create application profiles pretty easily --
> perhaps with menu-driven interfaces, and using their syntax
> of choice -- yet in a way that ensures that their data will
> play well as Linked Data.
>
> Asking people to conceptualize their data in terms of class
> semantics seems to set the barrier really high.  Explaining
> the simpler notion of a Dublin Core application profile has
> been challenging enough.  Should the design of good metadata
> require knowledge of OWL?  Can OWL semantics really be stuffed
> under the hood of a clever interface?


That would be the hope of course: I myself have sometimes trouble with OWL (well, I mean with the second version).
In fact I wouldn't mind keeping the current formulation of the AP stuff. But it would be a big plus if we could map under the hood this formulation to the SW tech stack, allowing interested implementors to benefit from the standardization status this stack has now gained.

Antoine

Received on Tuesday, 12 October 2010 18:13:42 UTC