Re: WordNet TF: short note on porting issues

Notes for today's telecon.
Concerning my actions:

At 17:29 +0200 31-03-2004, Guus Schreiber wrote:
>  ACTION: aldo describe Wordnet TF
>ACTION: aldo e-mail WordNet update to mailing list

wrt my outline, after a discussion with Jeremy Carroll (and others), there 
seems to be an agreement on splitting the task force into two (or three?) 
phases (the issues I listed at the end of the outline should be distributed/
modified accordingly):

(1) delivering an OWL-RDF schema for Princeton WordNet "as is", in order to 
port it to the SW. The basic port has two possible variants, which I describe 
below. Additionally, it would be helpful to produce a preliminary "general" 
schema that could be used for any wordnet

(2) investigating on basic improvements that can be made to Princeton WordNet 
in order to derive SW ontologies from it, and to produce recommendations for 
wordnets in general (possible joint work with thesauri task force)

(3) investigating on advanced improvements, in order to derive richer 
ontologies from wordnets, and produce suggestions/recommendations out of it.

Besides phases, an issue has emerged from looking at Dan Brickley's version of 
WordNet that deals with the management of large ontologies: the usefulness of 
maintaining large ontologies into as many 'views files' as many classes (or 
individuals?) are included in an ontology. 
I asked for existing tools to derive such 'views files', and Libby Miller 
suggests that this is a usecase for the Data Access Working Group.


>3. TASK FORCE TEMPLATE (10 min)

>Revised draft:
>http://lists.w3.org/Archives/Public/public-swbp-wg/2004JanMar/0208.html

>PROPOSED to accept the draft format.

I like the proposal, and will produce a first draft in that format before next 
telecon.


>4.2 WORDNET (15 min)

>Aldo's outline:
>http://lists.w3.org/Archives/Public/public-swbp-wg/2004JanMar/0114.html
>plus the thread following

>Discussion points:
>- scoping issues
>- short-term actions

See above. As far as short-term actions, I think we can conclude phase 1 soon, 
and deliver a recommendation for the schema and the related ontology. There is 
one issue to be resolved which concerns the variant of the port (see below).

>At 13:55 +0200 1-04-2004, Guus Schreiber wrote:
>I've produced a short note on porting issues for the various WordNet 
>constructs:

>  http://www.cs.vu.nl/~guus/public/wn.txt

>The note takes into account two existing ports (should be extended). The aim 
>is to be complete wrt the current WordNet distribution (could be subject of 
>discussion). It does not consider extensions.

>I hope it is useful input to reaching consensus on porting WordNet to RDF/OWL 
>format, hopefully also with the original developers in the loop.

Thanks, Guus! It is already a good thing indeed. 

I noticed that your schema is quite close to the one produced by the KID group 
for their OWL port.
Looking at the KID port, they made the schema, and then treated each element 
in the Wordnet database as an instance of some class in the schema.
This port is different in type from Dan Brickley's one, which directly assumed 
hyponymy relation as subClassOf.
These two strategies are what I call the two variants of WordNet porting as 
far as phase 1 is concerned.

Pros and cons:
a) the first strategy has the advantage of porting WordNet without changing 
anything in the original database, there just is a one-to-one mapping of (part 
of) the original schema into an OWL schema. The disadvantage is that one 
cannot exploit the subClassOf partial order. Moreover, in phase 2 would be 
more difficult to propose datatype changes.

b) the second strategy has the advantage of exploiting subClassOf, and being 
already in the direction of Phase2. On the other hand, the original schema of 
Wordnet is not maintained, since treating hyponymy as subClass is a semantic 
non-neutral interpretation.

I am in favour of b), despite its cons. In fact, looking at current research 
in wordnets and ontologies (including some of Princeton's), there seems to be 
a huge agreement in progressing towards the status of ontologies rather than 
staying at the level of lexical databases.
Moreover, in case b) is accepted, schemas like Guus' one can be used either as 
"meta-schemas", or be adapted to fulfil b).

BTW, a) is more or less done, and we could deliver a) solution as a first, 
minimal recommendation, with the obvious warnings.

Ciao
Aldo

-- 
Aldo Gangemi
Research Scientist
Laboratory for Applied Ontology
ISTC-CNR
Via Nomentana 56, Rome, Italy
+39.06.86090249

-------------------------------------------------
This mail sent through IMP: http://horde.org/imp/

Received on Thursday, 1 April 2004 11:53:28 UTC