W3C home > Mailing lists > Public > public-swbp-wg@w3.org > December 2005

RE: [WN] Fwd: WordNet Namespace

From: Peter Mika <pmika@cs.vu.nl>
Date: Thu, 15 Dec 2005 17:02:54 +0100
To: "'Mark van Assem'" <mark@cs.vu.nl>, "'Jacco van Ossenbruggen'" <Jacco.van.Ossenbruggen@cwi.nl>
Cc: "'Jeremy Carroll'" <jjc@hpl.hp.com>, "'Aldo Gangemi'" <aldo.gangemi@istc.cnr.it>, <public-swbp-wg@w3.org>, <schreiber@cs.vu.nl>, <Benjamin.Nguyen@inria.fr>
Message-ID: <002d01c60191$02f4c840$fb1e2582@fspc055>

Hi All,

I still think that word senses are the atomic concepts of WordNet and
excluding them altogether could hurt. So what I have in mind now is to
extend Mark's proposal by having three files:

1. one with synsets
2. one with labels for synsets
3. one that defines wordsenses, their relations to each other and their
relations to synsets

1+3 would make up what I consider the complete WordNet, but still people
could work with just 1 or 1+2. 

With regard to what the Princeton server should serve: in an ideal world how
to represent WordNet and how to serve it via the Web would be two separate
issues. But since higher powers have recently conferred us the idea that you
should be able to look up URIs, it is probably an issue. Since this is a
very recent development, it is hard to talk of a best practice here. See
also the discussion on what to server for SPARQL DESCRIBE queries [2].

Two alternatives may be:

1. All statements with this resource as subject, predicate or object.
2. The concise bounded description of the resource [1]

Cheers,
Peter

[1] http://www.w3.org/Submission/CBD/
[2] http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2005Nov/






> Hi all,
> 
> Before I comment I shortly summarize the discussion on the files:
> 
> 1- split WN-Full up into more convenient files
> 2- make a WN-Light
> 3- serve slash uris
> 
> 1) This is not a problem, will do so. Or, if we want slash URIs, this
> is not necessary because every resource becomes one file?
> 
> 2) I can create a separate file that attaches all WordSenses/Words as
> labels to their Synsets. The Synset file plus this new file then
> constitutes a WN-Light. You loose (a) Word and WordSense URI; and (b)
> the ability to use the relationships that are between WordSenses
> instead of between Synsets (pertainsTo, participleOf, seeAlso, antonym).
> 
> Peter Mika commented that taking out the WordSense-as-Class in favour
> of WordSenses-as-Labels seriously impairs an important use of WN,
> namely sense disambiguation. I don't know how WN is exactly used for
> sense disambiguation, but I can imagine this situation:
> 
> - program finds relevant word in a text
> - program searches Synsets with matching labels
> - program chooses between the returned Synsets
> - program annotates the word in the text with the WN Synset URI.
> 
> For this approach WN-Light is fine. Am I missing something or would
> WN-Light as I propose above be a good idea?
> 
> 3) I don't think we can ask Princeton to do something that is more
> complex than serving file(s) at a particular location (or can we,
> Aldo?). Is the slash solution practical enough?
> 
> If we do choose to use hash URIs, then should we have different
> namespaces for the different files? This also reduces the size of a
> download when you query for a WN URI.
> 
> Mark.
> 
> Jacco van Ossenbruggen wrote:
> > Peter Mika wrote:
> >
> >> Hi All,
> >>
> >> Regarding the second issue: I think it's less of a problem. Once you
> >> agree
> >> on the actual representation of WordNet, you can make it available
> >> either as
> >> a single file or a set of files, e.g. one for each resource,
> >> containing only
> >> the triples where that resource appears.
> >>
> > I agree, but you still need to define exactly what triples are returned
> > for each type of URI.
> > So, what subgraph does resolving http://wordnet.princeton.edu/rdf/entity
> > actually return?
> > And what about http://wordnet.princeton.edu/rdf/antonym?
> >
> >> The WordNet server can then either serve these static files or run a
> >> RESTful
> >> Web Service in the background that queries the ontology dynamically.
> >>
> >>
> > In theory, yes, but in practice I doubt that Princeton is going to
> > develop such a service.
> >
> >> In any case as Jeremy says what you want are slash URIs.
> >>
> >>
> > I agree.
> >
> > Jacco
> 
> --
>   Mark F.J. van Assem - Vrije Universiteit Amsterdam
>         markREMOVE@cs.vu.nl - http://www.cs.vu.nl/~mark
Received on Thursday, 15 December 2005 16:10:08 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 8 January 2008 14:17:19 GMT