[tm-pubsubj] Classes are singular nouns ... at least in ontologies

[Bernard Vatant <bernard.vatant@mondeca.com>]

This message is dedicated to Pete :)

Just to make a complete rationale about classes being written with
singular nouns, and why I don't think it's a detail, and how it
indirectly meets some other important current lines of thought

1. This is the recommended and usual current best practice in
ontologies. 
See e.g. the example developed in OWL guide
http://www.w3.org/TR/2002/WD-owl-guide-20021104/#SimpleClasses
See that classes here are named "Winery" "Region" and so on ...

2. Is that only in OWL, RDF and relative Description Logic land? 
Nope. Just a little story : in Mondeca applications, we used to keep the
plural for practical internal purposes, as a way to make distinct, e.g.,
the topic class "Products" vs the role type "product" in a "production
association". We hoped that nobody would care about that lexical
"bricolage". We were wrong. I had several times remarks about it, from
documentalists and thesaurus experts, that it was a bad practice and
that we should do otherwise. In French as well as in English, BTW. Those
people are very picky, and they care about that kind of thing(s).

3. It happens that documentalists, taxonomists, ontologists and the like
are the first (main, maybe only) target readers and users of our
specification. We want those folks to get on board. We've better write
something in their language rather than in street english. 

4. It has been discussed to use "group" or "set" instead. Bad idea.
People do not care about identification of groups or sets. They want to
identify classes and instances. Mainly classes, actually. The
"class-instance" paradigm is fundamental in ontologies, and ontologies
are here to stay, making a silent revolution in various industries. I
already wrote that lately in response to Thomas on topicmapmail, but I
insist on it. My view is that if we don't get people to use PSIs to
identify classes in ontologies, our specification will be close to
useless. In OWL, classes are identified by URIs. OWL classes will be IMO
the best candidates for PSIfication outthere. 

Bottom line - This leads to another issue, that we have only evoked so
far. If ontology classes are identified by PSIs, how do you ensure and
check consistency of class relationships and properties declared in
ontologies with relationships declared in topic maps using those classes
PSIs? This issue is fundamental and deserves certainly a specific thread
in a larger forum than this one, which is supposed to be "private" (see
forwarded exchange with Karl Best), and certainly input from ontology
wizzards.

More to come

Bernard

_____________________________________

Bernard Vatant
Senior Consultant - Knowledge Engineering
www.mondeca.com
_____________________________________

| -----Original Message-----
| From: larsga@pc36.avidiaasen.online.no
| [mailto:larsga@pc36.avidiaasen.online.no] On Behalf Of Lars Marius
Garshol
| Sent: jeudi 6 février 2003 10:38
| To: 'tm-pubsubj'
| 
| 
| * Bernard Vatant
| |
| | Agenda
| |
| | Discussion of Steve's re-drafting of
| | "Gentle Introduction to Published Subjects"
| | http://www.ontopia.net/tmp/pubsubj-gentle-intro.htm
| |
| | We have still a few hours to post preliminary comments here to
bootstrap
| | the discussion.
| 
| I sent Steve the comments below, but he didn't have time to apply
| them, due to other duties popping up at short notice, so I am
| reposting them here for the benefit of everybody.
| 
| ----------------------------------------------------------------------
| 
| I think the intro uses pretty much the right tone. It's semi-technical
| yet pretty gentle, and I think it's quite good. I don't think it
| covers all the things people will be wondering about, but we can
| extend that later.
| 
| Some specific comments:
| 
| 2.1: RDF "resource" is equivalent to TM "subject", not "topic"
| 
| 2.2: "Subject identity". WG3 decided to discard this term as an
|      official term, and so the SAM uses "Subject identification"
|      instead. I feel that's a better term, because *that's* really
|      what it is about. (I don't feel this needs to be a formal term
|      that is introduced and defined, though. Presenting the issue is
|      enough, I think, and we shouldn't think of this as a formal
|      term, IMHO.)
| 
| 2.3: "through the use of URIs": i.e, addressing with URIs
| 
|      I think a diagram here is a good idea, as it makes for a good
|      contrast with the diagram below
| 
|      SAM does *not* use the terms addressable subject and
|      non-addressable subject. See
|      <URL: http://www.isotopicmaps.org/sam/sam-model/#d0e582 > and
|      <URL: http://www.isotopicmaps.org/sam/sam-model/#d0e315 >
| 
| 2.3: you have two 2.3 sections :-)
| 
| 2.3.1: "network-retrievable information resource": there is no other
|        kind, so this is like a rider on horseback
| 
|        re NOTE: the subject indicator is an information resource,
|        which may of course be regarded as a subject by creating a
|        topic with a subject address. Probably worth explaining this
|        explicitly and relating it to the diagram above.
| 
| 2.3.3: the example definition is problematic because it actually
|        defines four different subjects. I think the intro will have to
|        relate to that problem somehow. (I see now that you are aware
|        of it, but I think you need to flag this problem more clearly.
|        Some people will read this without realizing the problem, I
|        think. Perhaps listing the four subjects more explicitly in
|        2.4.1 would help.)
| 
| 2.4.2: "availaility" :)
| 
| 2.4.3: another benefit is the higher precision of the definition
| 
| --
| Lars Marius Garshol, Ontopian         <URL: http://www.ontopia.net >
| GSM: +47 98 21 55 50                  <URL: http://www.garshol.priv.no
>
| 
| 
| ----------------------------------------------------------------
| To subscribe or unsubscribe from this elist use the subscription
| manager: <http://lists.oasis-open.org/ob/adm.pl>