$)Cvcr and television (was RE: WordNet in RDF/XML: 50,000+ RDF class vocabulary...)

http://snowball.ilrt.bris.ac.uk/xmlns/wordnet/noun/appliance
http://snowball.ilrt.bris.ac.uk/xmlns/wordnet/noun/vcr

http://snowball.ilrt.bris.ac.uk/xmlns/wordnet/noun/tv
http://snowball.ilrt.bris.ac.uk/xmlns/wordnet/noun/television
http://snowball.ilrt.bris.ac.uk/xmlns/wordnet/noun/television_tube
http://snowball.ilrt.bris.ac.uk/xmlns/wordnet/noun/crt

http://snowball.ilrt.bris.ac.uk/xmlns/wordnet/noun/software
http://snowball.ilrt.bris.ac.uk/xmlns/wordnet/noun/fortran

I tried the above and find a very little useful association between eg:
-tv and television
-televison tube (or crt) and television (or tv)
-software and fortran

I was expecting better eg:
-vcr and tv to show as related electronic devices or appliances
-television tube to show as an electronic part of the television
-fortran to be under software
-tv an alternative name for television (other than in comments)

Maybe I was expecting too much?
Richard Humpleman,
SISA, Samsung Electronics,
San Jose, CA.


>>> -----Original Message-----
>>> From: Dan Brickley [mailto:danbri@w3.org]
>>> Sent: Thursday, December 09, 1999 4:39 PM
>>> To: www-rdf-interest@w3.org
>>> Subject: RE: WordNet in RDF/XML: 50,000+ RDF class vocabulary...
>>> 
>>> 
>>> On Thu, 9 Dec 1999, Richard Humpleman - SISA wrote:
>>> > This seems to have stopped working but it worked great 
>>> before. Any chance
>>> > to get it going for a little while looonger?
>>> > 
>>> > Reference to undeclared namespace prefix: 'r'. Line 14, 
>>> Position 51 
>>> > <Class r:about="http://xmlns.com/wordnet/1.6/cat
>>> 
>>> Ooops, thanks. I've fixed it.
>>> 
>>> I had a number of comments about the unwieldy long URLs I 
>>> was using, and
>>> the multiple levels of namespace management they depend 
>>> upon. So I moved
>>> it to a http://xmlns.com/wordnet/1.6/* address instead, 
>>> which is a domain
>>> explicitly committed to (eventual) persistence and 
>>> non-cheesiness. In
>>> process, I broke the script and a bunch of URLs -- so much 
>>> for persistence!
>>> 
>>> 
>>> Ideally, Princeton would host the canonical URL for WordNet 
>>> since they
>>> manage the vocabulary. Failing that, I think there's a lot 
>>> to be gained by
>>> agreeing within the RDF community a single namespace URI 
>>> for WordNet 1.6
>>> concepts. Using purl.org or something under desire.org are other
>>> possibilities I've been toying with.
>>> 
>>> <aside>
>>> I have no personal desire to manage xmlns.com for all time so am
>>> looking for ways of offloading the domain to a committee of 
>>> do-gooder
>>> RDF/XML enthusiasts who want short reliable names for 'semantic web'
>>> namespaces. Neither ILRT, University of Bristol nor W3C have any
>>> commitment to maintaining names in that namespace. So - I'd 
>>> like to hand
>>> off reponsibility for xmlns.com to some entity more reliable than
>>> myself. Suggestions (in separate thread) welcomed :-)
>>> </aside>  
>>> 
>>> Anyway, regarding WordNet I need to flag up a major issue: 
>>> my current demo
>>> conflates 'word senses' with the words associated with 
>>> those senses. My
>>> rather dusty knowledge of WordNet is that 'senses' are 
>>> clusters of broadly
>>> equivalent terms, eg. as shown comma-separated here:
>>> 
>>> > >>> 	tree
>>> > >>> 	       => woody plant, ligneous plant
>>> > >>> 	           => vascular plant, tracheophyte
>>> > >>> 	               => plant, flora, plant life
>>> 
>>> So the sub-class relation is between clusters of terms. 
>>> This leaves us
>>> with a dillema: do we assign URIs to senses, terms or both? 
>>> I'd expect
>>> both, but with most classifications happening in terms of 
>>> senses. In which
>>> case the question is -- what identifier is appropriate to 
>>> id a word sense?
>>> 
>>> I'm hoping that someone more familiar with details of 
>>> WordNet will step
>>> in and tell us how best to do this... Suggestions anyone?
>>> 
>>> Dan
>>> 
>>> > 
>>> > Regards,
>>> > Richard Humpleman.

-----cut-----

Received on Thursday, 9 December 1999 20:24:22 UTC