RE: SUO: RE: Re: Comment #13 - 'NL'
> John,
> I agree with this proposal with one change. I'd suggest that we not =
> limit p
> to being predicates. It will also be valuable to include classes and
> instances along with their mapping to a lexicon and tagging as a part =
> of
> speech.
> Adam
p ALREADY INCLUDES classes and instances which, IMHO can be modelled as a unary
predicate. Any logicians out there to confirm this?
Mike
----
From graham.horn@aihw.gov.au Tue Sep 26 01:16:22 2000
From: "Horn, Graham" <graham.horn@aihw.gov.au>
To: "'apease'" <apease@teknowledge.com>, sowa@bestweb.net,
mfu@redwood.rt.cs.boeing.com, standard-upper-ontology@ieee.org
Subject: RE: SUO: RE: Re: Comment #13 - 'NL'
Date: Tue, 26 Sep 2000 19:14:40 +1000
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Hi all,=20
. I support Adam's refinement to John's approach and proposal, here
for the Standard Upper Ontology. The resulting definition supports =
clarity
and unambiguity with minimal restrictions to word usage, and is in line =
with
what I understand as the main principles underlying the word "concept". =
=20
. In general, I would suggest preference for use of ordinary English
words and meanings, wherever they are clear and unambiguous, without =
resort
to the sort of narrow definition applied in this case. This is because =
the
need to resort to a set of definitions of terms reduces both the =
accuracy of
interpretation by the public at large, and also people's ability to be
confident of understanding the standard without constant reference to =
the
set of definitions of terms. However, I am happy that concepts =
important to
the project, such as this one, carry the further restriction in the
interests of ensuring precision.=20
. By the way, we have just posted the new version of my main project
here, AIHW's "Knowledgebase" on our website at
http://www.aihw.gov.au/inet/knowledgebase/index.html. Any comments are
welcome.=20
Cheers Graham Horn
Australian Institute of Health and Welfare=20
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
Phone: 02.6244.1094 =20
Fax: 02.6244.1199 =20
E=ADmail: Graham.Horn@aihw.gov.au <mailto:graham.horn@aihw.gov.au>
-----Original Message-----
From: apease [mailto:apease@teknowledge.com]
Sent: Tuesday, September 26, 2000 4:00 PM
To: sowa@bestweb.net; mfu@redwood.rt.cs.boeing.com;
graham.horn@aihw.gov.au; standard-upper-ontology@ieee.org;
joe.christensen@aihw.gov.au
Subject: Re: SUO: RE: Re: Comment #13 - 'NL'
John,
I agree with this proposal with one change. I'd suggest that we not =
limit p
to being predicates. It will also be valuable to include classes and
instances along with their mapping to a lexicon and tagging as a part =
of
speech.
Adam
At 09:50 PM 9/25/2000 -0400, John F. Sowa wrote:
>Mike,
>
>As you mentioned in an earlier note (and which I supported), a good =
way to
avoid hangups about particular words is to edict that they shall be =
avoided.
Another way is to define them in a very narrow technical sense and to =
forbid
any other possible usage among project participants. I propose that we
adopt the latter technique for the word "concept".
>
> >One key difference between an ontology and a natural language
lexicon is that the former is *concept* focussed, and the latter is =
*word*
or *term* focussed.
>
> >What is primal in an ontology are the concepts in a domain. Terms
are secondary, but of course very important to have. They refer to the
concepts.
>
> >What is primal in a natural language lexicon are *words*, or
*terms*. These words have meaning, but that is not what is intended to =
be
captured and represented in a natural language lexicon (NLL). Instead, =
an
NLL is intended to capture the key relationships between the words =
insofar
as how they occur in sentences, as well as how their meaning is similar =
to
other words.
>
>Since the word "concept" seems to be a fairly natural word to use in =
many
of our discussions, I suggest that we keep the word, but define the =
term
"SUO concept" in the following way:
>
>Definition: An _SUO concept_ c is a triple (p,t,s), in which p is a
predicate defined by a definition or axioms in KIF, t is an English =
term
(word or multiword phrase), and s is an English syntactic category
represented by one of the following character strings: "noun",
"intransitive verb", "transitive verb", "ditransitive verb", =
"adjective",
"adverb", "preposition", "coordinate conjunction", "subordinate
conjunction", or other strings that may be specified by the SUO.
>
>Guideline for SUO documents: When an SUO concept c is mentioned in =
any SUO
document, the formal meaning of c is determined by the KIF predicate p =
that
forms the first part of the triple of c. In any English commentary,
paraphrase, or explanation, c may be expressed by the term t used in =
the
syntactic category s according to the commonly accepted rules of =
English
grammar.
>
>Since this proposal allows the predicate p to have any valence =
(monadic,
dyadic, triadic, ...), it is general enough to include relations as =
well as
simple terms. It also allows the SUO to support languages other than
English by leaving the predicate p of a concept (p,t,s), but =
translating t
to some other NL word or phrase and translating s to the corresponding
syntax.
>
>Note: this option allows a concept expressed by a one-word term in =
one NL
to have a multiword term in another NL, but both NL's would defer to =
the KIF
predicate as the "meaning".=20
>
>John
-----------------
Adam Pease
Teknowledge
(650) 424-0500 x571