Hmmm. Shades of email lists long long ago. (01)
As for "concept" and your triangle, this is a more general concept than
class or type. Typing and/or classifying things is a kind of concept
(trying not to get strung up by my own words here). In other lexicons this
seems more like "statement" or "fact". I find "fact" to assertive (many
statements are, at best, opinions). Statement assumes a speech act - also a
valuable but more specific concept. A term is also a concept! So yes,
concept would seem a reasonable name for an atom of information. (02)
Given class Vs. Type I would much prefer type, perhaps this is due to my
information systems background where a class is a factory for instances, not
just a kind of instance. Type seems the more general and more precise
concept with a good match to human speak. I thought this was getting
accepted, then OWL came out. (03)
-Cory (04)
-----Original Message-----
From: ontac-dev-bounces@xxxxxxxxxxxxxx
[mailto:ontac-dev-bounces@xxxxxxxxxxxxxx] On Behalf Of Cassidy, Patrick J.
Sent: Wednesday, January 18, 2006 11:26 AM
To: ontac-dev@xxxxxxxxxxxxxx
Subject: [ontac-dev] Type vs. Class et al. (05)
COSMO-WG:
This is a first note using the ONTAC-dev listserver. I did not
receive any objections to moving our discussions to the ONTAC-dev
server, so all COSMO-WG members should find themselves on the list. (06)
Regarding the issue of what to call a 'Class' - Type? . . . (07)
First a clarification: (08)
>> The first [concept] is a disaster, though it is used e.g.
by ISO in its current terminology standards, and indeed by our Great
Leader. (09)
[PC] modestly accepting what I believe is a reference to my frequent
use of the term "concept", please understand that I fully sympathize
with Barry's emphasis on the need to distinguish between real things in
the world and the concepts people have of them and the terms by which
we refer to them. But I am not aware yet of a usage that will serve
all of our purposes, and have noticed some problems caused by a lack of
standardization of these terms. So it's probably a good idea to try to
agree on a terminology. (010)
I need a term that means "a class or relation or function or instance
or metaclass or function term or axiom or procedural rule or attached
method or any other element that we would want to put into an ontology
that represents some element of meaning, as distinct from the actual
things in the world to which they refer and as distinct from the
specific manner in which they happen to be represented." or "some
abstract entity which is represented by symbols in our ontologies, is
intended to correspond in structure to some idea people have about
something or other, and refers to something other than itself". (011)
For that, I have been using the term "concept". Occasionally in
informal notes I may use "notion". In deference to Barry's dislike of
'concept', I have occasionally used the phrase "Class or relation",
though that leaves out all those other things I want to refer to. (012)
I think this is close to what is usually intended by references to
Ogden's "meaning triangle": (013)
Concept
/ \
/ \
/ \
Symbol/Term --- Referent/Thing/Real-world Object (014)
Does anyone else have a candidate for a general term we should use to
refer to any of the constituent elements of our ontologies? (015)
----
When I mean a "Class" (Type, Category, Kind) I say "Class". (016)
I use "Class" because it is used by the RDF and OWL communities (which
are W3C standards), and I try to use the terms which are most widely
used. I believe this to be identical in meaning to "Collection" as
used in OpenCyc, "Class" in SUMO, "Universal" in DOLCE and "Property"
in the Ontology Works system. I use it in the Ontolingua sense, as an
intensionally defined grouping distinct from 'Set' which refers to an
extensionally defined grouping. I am not intending to use it in the
set-theoretic sense. (017)
That term may have different usages in different communities (I think
ISO15926 has a slightly different usage, though I do not yet fully
understand it). (018)
For our discussions, I will happily use any term on which we can agree. (019)
As for Barry's list: (020)
[BS] >> There is no perfect solution here. The
conceivably feasible alternatives (known to me) are: (021)
concept
universal
type
class
species
kind
category (022)
[PC] . . . you left out "Sort" and "Property" (the latter from Ontology
Works). (023)
One problem I struggle with in trying to be precise is that there are
several relations labeled 'refers-to' (or 'represents'). Whatever we
have in our heads (a mental process or mental object?) 'refers-to'
things in the real world. An abstract notion in our ontologies
'refers-to' those things we have in our heads. The symbols we use to
communicate 'refer-to' the abstract notions as well as the things in
our heads, and also to the things in the real world. Are those
'abstract notions' the same as Ogden's 'concepts' or not? (024)
For our immediate purpose we only need to fix on the term we will use
to refer to those intensionally defined groupings called: (025)
Class in Ontolingua and Protege
Class in RDF and OWL
Class in SUMO
Collection in OpenCyc
Universal in DOLCE
Property in Ontology Works' IODE system
--------------- (026)
The vote is still open. (027)
Pat (028)
p.s. I prefer 'Beloved Leader' ;-) (029)
Patrick Cassidy
MITRE Corporation
260 Industrial Way
Eatontown, NJ 07724
Mail Stop: MNJE
Phone: 732-578-6340
Cell: 908-565-4053
Fax: 732-578-6012
Email: pcassidy@xxxxxxxxx (030)
======== Original Note 2005-01-18 ============ (031)
>Chris Menzel:>
>Re moving to "type", for sheer ambiguity, "type" is no better than
>"class"; maybe even worse. It has at least four common senses:
>
>1. The roughly Aristotelian sense in question, i.e., an "intensional
>category".
>
>2. A logical sense indicating the level of an entity in the theory of
>types (of which there are two flavors, simple and ramified, which
>themselves come in several varieties, e.g., Russell's simple theory
vs.
>Montague's).
>
>3. A corresponding linguistic sense indicating formal grammatical
>category corresponding to a semantic type.
>
>4. A programming sense, viz., a datatype, along with a corresponding
>syntactic sense for variables that are declared to be of one type or
>another, indicating that their range is restricted to the instances of
>that type.
>
>Of course, as with class, we could just axiomatize the sense of "type"
>we want. But my point here is that, since "type" is no less ambiguous
>than "class", there's no reason not to stick with "class". (032)
There is no perfect solution here. The
conceivably feasible alternatives (known to me) are: (033)
concept
universal
type
class
species
kind
category (034)
The first is a disaster, though it is used e.g.
by ISO in its current terminology standards, and indeed by our Great
Leader. (035)
The second has the advantage that it means
exactly what we want to say, though it comes with
philosophical baggage which some might not like,
and it is not part of the lexicon of ordinary English. (036)
The third ('type') has the advantage that it IS
lexicalized in ordinary English to mean just what
we want to say. Moreover, against Chris and Cyc
(and with Leo and the English language) it has
the advantage that it allows us to make a clear
distinction between type (meaning: the entity
which gets instantiated, as in 'Fido instantiates
the type dog') and class (meaning the extension
of this entity, the set or totality or collection of instances). (037)
The third has the further advantage that it looks
as if there is a real possibility that ISO might
adopt it to clarify current confusions surrounding 'concept': (038)
http://ontology.buffalo.edu/concepts/ConceptsandOntologies.pdf (039)
If this goes through, 'concept' will mean, for
ISO: 'meaning of a general term'. (040)
The fourth ('class') is needed to mean: extension of a type. (041)
The fifth and sixth ('species', 'kind') probably
will not generalize well beyond biology, though
the latter, especially, has some of the advantages of 'type'. (042)
The sixth ('category') we need (à la Aristotle) for 'highest type'. (043)
---------
Regarding Chris's point that we should not be worrying about
terminology: (044)
I think Chris is forgetting that ontologies --
and particularly an ontology of the sort
envisaged by Pat/ONTAC -- must point in two
directions: towards humans, who will be involved
in conceiving, and using, and curating it; and
towards machines and logicians, who, as Chris
rightly points out, could care less about lexical
tags. Humans need to be led by the hand, both by
good documentation and by friendly principles
governing choice of terms. Otherwise an ontology
will have no chance of being widely used, and of
being attractive enough, and intelligible enough,
that systems will be built in its terms and
thereby become 'interoperable' with its help.
(Ontologies are, in this respect, like
telephones: if you do not have a wide network of
users, then your ontology will do nothing for
you.) (For this reason also, of course, ontologies should not be kept
secret.) (045)
_________________________________________________________________
Message Archives: http://colab.cim3.net/forum/ontac-dev/
To Post: mailto:ontac-dev@xxxxxxxxxxxxxx
Subscribe/Unsubscribe/Config:
http://colab.cim3.net/mailman/listinfo/ontac-dev/
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki:
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG (046)
_________________________________________________________________
Message Archives: http://colab.cim3.net/forum/ontac-dev/
To Post: mailto:ontac-dev@xxxxxxxxxxxxxx
Subscribe/Unsubscribe/Config: http://colab.cim3.net/mailman/listinfo/ontac-dev/
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki:
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG (047)
|