[WN] Wordnet review

Dear all,

Per my action item:
http://www.w3.org/2006/02/06-swbp-minutes.html#action16

I provide a brief 2nd review of:
http://www.w3.org/2001/sw/BestPractices/WNET/wn-conversion.html 
(Editor's Draft 2 February 2006)
below.

Regards, Jacco

--
General remarks:

First, I've seen no systematic reply from the editors to the issues 
raised in the first round of reviews.  I've not noticed any omissions, 
but did not do a systematic check myself either.

Second, there are still a lot of [TODOs] left, and I found that the many 
short sections and many appendices still make it a quite a hard read.

Having said this, I think this current document is much, much better 
that the previous version I reviewed for the f2f in Galway.

The main thing I miss are some basic guidelines on "how do I use this as 
an annotator".  I am a bit reluctant to bring this up because it might 
get us into the httpRange-14 discussion again, I'd I would prefer to 
avoid that.  It may also be out of scope to go into this deeply, but 
some simple answer to basic questions like: if I want to annotate some 
resource as having a myns:color property with value wn:red,  do I use 
the URI to the Word, the WordSense or the Synset or 'red'? If I want to 
use one of those options here, what are the typical use cases for the 
other for refering to the other instances? And what if I want to make a 
claim that I disagree with the definition of Princeton for 'red' in Wordnet?

I would also appreciate it if the new instance date would become 
available.  During the first review, many issues only came up after 
using the rdf files in a real application (Mark knows which app I'm 
talking about:-).
 
Small remarks:

Section 1.
       Princeton hosts the conversion of the most recent version of 
WordNet RDF/OWL at the following URI:
       http://wordnet.princeton.edu/wn/

 - Is there agreement with Princeton about hosting the RDF and the URI?

Appendix F: Relation to previous versions
The discussion about the Word 'chat' suggest that all synsets are 
language independent.
I'm no expert, but it looks to my like an assumption that may be 
questionable, and certainly not one to make implicitly.
Maybe you could have a single (sub)section devoted to i18n?
This could also discuss what URIs to use for other languages, how to 
link related intances of Word, WordSense and Synset of different 
languages, etc. The use of diacritics etc.
What about the use of xml:lang='en' .  Do you need to distinguish 
between US and UK English here?  (I think you do!).
All i18n stuff is now scattered over the document.

 Appendix H: Open Issues
 - Need to check for possible bugs in Prolog conversion
   program that make it generate wrong RDF output. Also use
   DL reasoner to check for problems.

I'm no expert in this, but isn't the mere fact that you use rdfs:label 
on instances sufficient to move you out of DL?
Could this be a reason not to use rdfs:label in for Word instances?

Received on Wednesday, 15 February 2006 16:22:36 UTC