Re: SKOS vs OWL

From: Ed Shaya <eshaya-at-umd.edu>
Date: Mon, 01 Oct 2007 15:21:19 -0400


Norman,

        The proof of the pudding is in the eating. You are right that skos treats the terms as instances and that does not mesh great with normal ontologies. But that does not mean that a skos/owl presents any difficulties or confusions once you see one. So, I looked around for a skos file and found this BLOG Housekeeping "http://norman.walsh.name/knows/taxonomy" and I converted it to OWL. I admit that I did most of it in vim, but I used Protege and WonderWeb to check validity. Since the terms are Individuals the only new owl:Class is Topic which I made a subclass of Concept. It seems to me that this should have been explicit in the skos as well, but it was not. You can compare, except for some Walsh icons and anchors, this pretty well faithfully reproduces the skos. There were three or so terms Application, IndexTerm, and LinkGroup which were rdfs:Class. I could have just made these owl:Class, but I thought they were also instances of Concept so I did that. The point here is that now you can use OWL tools as RDF editors for skos. Since this is mostly Instances, then one may want to Configure to have the InstanceTree tab.

All of this disregards my actual feelings on this. And that is that we should really be interested in advancing automated search and automated search needs to know true subclassing. If I search for a spiral galaxy, the machine needs to know that an Sa galaxy is acceptable and a spiral arm is not. In skos they are both just narrower.

Ed

Norman Gray wrote:
>
> On 2007 Sep 30, at 15:44, Norman Gray wrote:
>

>> But then you're developing a vocabulary and an ontology both at once, 
>> and it's not clear (to me) what work the ontology half is doing.

>
> Reviewing again the recent messages here, I realise Ed and Tony did
> address this.
>
> Ed (2007 September 25):
>
>> it becomes both an OWL document and a skos document, which means that 
>> you can use all of the owl editors and diagramming tools.    I think 
>> this goes under the category of having your cake and eating too.

>
> Unfortunately, I think that cake would be fattening, and not
> particularly nourishing.
>
> The mechanism I'd propose for creating vocabularies would be to simply
> create SKOS normalisations of existing vocabularies (IAU, A&A, and so
> on); thus the SKOS would be created by scripts semi-mechanically,
> removing the need for editors or diagrams.
>
> Tony (2007 September 27), specifically addressing the
> acceleration/kinematics issue:
>
>> So we have ontological classes
>> vont:acceleration and vont:kinematics which each has vont:definedBy
>> relationships to vocab:acceleration and vocab:kinematics [...] existing
>> within their own hierarchies [...] but accessible from the same 
>> structure so that
>> vocab-based and ontology-based lookups can happen with the same object 
>> (and
>> across structures where appropriate)

>
> It's certainly true that, if there were work to be done to create an
> ontology of this type, then it would be natural to make links to the
> corresponding vocabulary. But that doesn't mean you'd have to do both
> jobs simultaneously: (a) the appearance of both vont:Star and vocab:Star
> would be confusing to anyone coming to this fresh, who'd have to be told
> the whole Concept vs. Class vs. instance story, and (b) OWL is just
> _bigger_, in its conceptual baggage, in its defining documents, and in
> its applications.
>
> In both cases, I don't believe the OWL aspect comes for free (contra
> Ed), and I don't see the requirements that oblige us to bear the cost.
> Though I'd be interested to be proved wrong.
>
> All the best,
>
> Norman
>
>
> --------------------------------------------------------------
> Norman Gray : http://nxg.me.uk
> eurovotech.org : University of Leicester, UK
>
>

Received on 2007-10-01Z21:21:40