-----Oorspronkelijk bericht-----
Van: Robert Spillers <skydog@pacbell.net>
Aan: Piek Vossen <piek.vossen@let.uva.nl>
CC: John F. Sowa <sowa@west.poly.edu>; jmc@cs.Stanford.EDU
<jmc@cs.Stanford.EDU>; lauspitz@world.std.com <lauspitz@world.std.com>;
Adam_Farquhar@KSL.Stanford.EDU <Adam_Farquhar@KSL.Stanford.EDU>;
Andreas.Reuter@eml.villa-bosch.de <Andreas.Reuter@eml.villa-bosch.de>;
Klaus.Tschira@ktf.villa-bosch.de <Klaus.Tschira@ktf.villa-bosch.de>;
doug@csi.uottawa.ca <doug@csi.uottawa.ca>; feigenbaum@KSL.Stanford.EDU
<feigenbaum@KSL.Stanford.EDU>; fellbaum@clarity.princeton.edu
<fellbaum@clarity.princeton.edu>; fritz@cyc.com <fritz@cyc.com>;
geo@clarity.princeton.edu <geo@clarity.princeton.edu>;
guarino@ladseb.pd.cnr.it <guarino@ladseb.pd.cnr.it>; hovy@isi.edu
<hovy@isi.edu>; jamesp@cs.brandeis.edu <jamesp@cs.brandeis.edu>;
lee@sabre.org <lee@sabre.org>; lee@textwise.com <lee@textwise.com>;
onto-std@KSL.Stanford.EDU <onto-std@KSL.Stanford.EDU>;
p.m.simons@leeds.ac.uk <p.m.simons@leeds.ac.uk>; peters@csli.stanford.edu
<peters@csli.stanford.edu>; phayes@coginst.uwf.edu <phayes@coginst.uwf.edu>;
phayes@nuts.coginst.uwf.edu <phayes@nuts.coginst.uwf.edu>;
polanyi@pal.xerox.com <polanyi@pal.xerox.com>; tsujii@is.s.u-tokyo.ac.jp
<tsujii@is.s.u-tokyo.ac.jp>; vdberg@us.ibm.com <vdberg@us.ibm.com>;
wahlster@dfki.uni-sb.de <wahlster@dfki.uni-sb.de>
Datum: vrijdag 8 mei 1998 8:40
Onderwerp: Re: Hayes' scrutiny of Top 40
>I agree. The lexicon is an excellent idea.
>
>So... what features should it have? How will it interface with an
ontology.
>What function(s) should be emphasized - information retrieval, machine
>translation, inferencing, etc? Can it be tuned to a specific purpose - by
what
>technique?
>
>Bob
>
>Piek Vossen wrote:
>
>> >I would put the issue Lee has raised on the list of topics to be
>> >addressed in Heidelberg. I think that we will have to draw some sort
>> >of distinction between the words used in a natural language and the
>> >categories used in one of our ontologies.
>> >
>> >I firmly believe that they are strongly related, but I also think
>> >that we should use a lexicon to provide the mappings between them.
>> I agree with this. It will be one of the key issues to be discussed. One
of
>> the things which makes wordnets so odd at the higher level, is that the
>> principle to relate word to their hyperonym no longer works above a
>> particular level (let's say above the cognitive basic level of
>> categorization (Rosch 1977)). Beyond that level you need something else
to
>> capture inferences and different types of hyperonyms to create a wordnet
>> that explains word-substitutions that can be used in Information
Retrieval,
>> Language Generation, etc.
>>
>> Piek.
>
>
>