Re: Announcement: RDF Representation of Wordnet available for Download

On Sat, 3 Feb 2001, Aaron Swartz wrote:

> Stefan Decker <stefan@db.stanford.edu> wrote:
>
> > an RDF-Representation of Wordnet 1.6 (a Lexical Database for English)
> > is available for download at:
> > http://www.SemanticWeb.org/library#wordnet
>
> Stefan / Dan,
>
> What is the relationship between this and the stuff at:
>
> http://xmlns.com/wordnet/1.6/

WordNet as an RDF Schema? An idea whose time has come...

I announced the predecessor to this in dec '99, though have only recently
figured out how to work around a problem with mapping terms to classes.
http://lists.w3.org/Archives/Public/www-rdf-interest/1999Dec/0002.html


xmlns.com is a domain I set up after I noticed folk starting to use the
long and fragile testbed URIs I set up.

The initial goal is to map all the noun terms from WordNet into
an RDF XML-namespace, such that the words become XML qualified names, and
can be used as typedNodes in the RDF XML grammar.



For eg.,

<wn:Organisation foaf:homepage="....">
 <util:employs>
   <wn:Person foaf:mbox="...">
     <util:depicted>
        <wn:Photograph about="..."> etc etc.

ie. bootstrap the pedantic web with thousands of useful classes for common
categories. It's not the most beautiful ontology ever created, but its
pretty good for many apps, and is freely available.

This representation is slightly different to the model Stefan and Sergey adopt.

The main difference relates to an outstanding bug in my implementation:
the relationship beteen synonym-sets and the terms that make up the
synset.

My goal is to allow the terms from the synset to each be used as an RDF
class. Currently I mess this up; what I've decided to do is declare a
class for each term and make a trivial superclass corresponding to the
synset. The class will have a generated ID. People will mostly use classes
of the form http://xmlns.com/wordnet/1.6/Xyz_n ie. initial capitals for
class name (term being 'xyz') and a numeric suffix for sense number.

A bit of a hack but the only way I can think of having a reasonable class
hierarcy and still projecting into a nice XML namespace for use in
instance data. Each class name should (and nearly does for xmlns.com)
dererence to a chunk of RDF data describing that class and others nearby
in the class hierarchy.


> Any best practices for use, etc?

Those of us who've been working with Wordnet in RDF should probably get
our heads together and try to come out with a common approach. I'd be
happy to offer the use of the xmlns.com namespace for such a collaborative
effort; it's nice and short and looks good in instance data.

For other RDF/Wordnet refs, see
http://www.google.com/search?q=wordnet+in+rdf

Dan

Received on Saturday, 3 February 2001 21:36:53 UTC