W3C home > Mailing lists > Public > www-rdf-interest@w3.org > January 2004

RE: [ontolog-forum] Wordnet Representations - XSD/RDF/OWL

From: Danny Ayers <danny666@virgilio.it>
Date: Fri, 23 Jan 2004 12:19:51 +0100
To: "[ontolog-forum] " <ontolog-forum@ontolog.cim3.net>
Cc: <www-rdf-interest@w3.org>
Message-ID: <BKELLDAGKABIOCHDFDBPIELLFDAA.danny666@virgilio.it>

(cc'ed to rdf-interest again)

re. http://www.daconta.net/project_folder/WordnetMetamodel.html

> On OWL DL versus RDF.  The current version which is parseable in RDF 
> does not use any of the rdfs vocabularly.  So, what I am asking is what
> do you feel the benefits of modeling the wordnet constructs as classes 
> as compared to the current resource-based approach.  What does saying a
> Concept (aka synset) is a Class buy us?

Ok, starting from the data available already, e.g.

<wn:Word ...>
	<wn:partsOfSpeech rdf:parseType="Collection">
		<wn:PartOfSpeech wn:type="verb">
			<wn:concepts rdf:parseType="Collection">
				<wn:Concept rdf:ID="_1943890">

the (striped) parsing of the syntax that's saying RDF interpretation stuff like

<_1943890> rdf:type Concept

and I /think/ there's an RDFS interpretation available as well from entailments, including bits like

Concept rdf:type rdfs:Class

and on top of that you could also apply an OWL Full interpretation...
But whatever, it's all as unconstrained as it could be. Your data example does look quite set-oriented (lots of collections) and I'm only guessing, but I think it should be possible to apply a load of OWL constraints. 

I nearly forgot - the benefit of the constaints being that it's easier/more efficient to query. You could probably bung the data into a RDBMS quite neatly, but presumably you want to use it alongside other ontologies and interfacing there could get messy. So using a common query/inference engine would be desirable. 

The benefit of using OWL DL is that it's decidable, or so the theory gos. I suspect that the practice at this point in time is that you're probably more likely to find a usable OWL DL engine than one for OWL Full. (See [1]) 

What is the cost? Hard to say. I don't think there would be any cost in the modelling, as I don't think you'll need to mix instances and classes in the way DL doesn't like. If this was something like RSS we were talking about, with loads of tools deployed that see RDF/XML as a bunch of regular expressions to be pretty printed, there would be a major cost. But (praise be!) it isn't, the slate is relatively clean. (Having said that, isn't there a DAML wn  interpretation somewhere, and didn't the good Mr. Brickley have a server for the terms too?)

But as I mentioned earlier, I reckon it would be easier to change from OWL DL -> Full (i.e. plain RDFS) if necessary after initial deployment than vice versa, the main reason being owl:Class is a subclass of rdfs:Class.
One thing I've been meaning to bring up on rdf-interest or somesuch is the idea of producing two versions of vocabularies - one geared towards OWL, one towards plain RDF. This could presumably cause complications if deployed in the wild on the Semantic Web, but at this point in time it may be easier to mix and match with other vocabs using RDFS publicly, but easier to reason with using OWL DL locally.

Couple of cents anyway.


[1] http://www.w3.org/2001/sw/WebOnt/impls
Received on Friday, 23 January 2004 06:29:28 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 15:07:49 UTC