[sc34wg3] Topic names and occurrences in the TMDM

Kal Ahmed sc34wg3@isotopicmaps.org
Fri, 30 Sep 2005 10:31:32 +0100

Hi Murray, Jan

I think that we have two separate but related questions here. To paraphrase
(and correct me if I am wrong)

1) Murray asks why a topic name can be treated as a specialized form of a
topic occurrence.
2) Jan asks why, if a topic name can be treated as a specialized form of a
topic occurrence, we need two separate concepts in the data model.

To address Murray's question first. In my opinion, topic occurrences are
used to convey "information related to the topic". The name "occurrence" is
mis-leading because it makes most English-speakers think that a topic
occurrence can only point to where the topic occurs in resources. So, a
topic occurrence on the topic "Kal" could be of type "CV" and point (via a
URI value) to some suitable HTML resource on the web. Equally, a topic
occurrence on the topic "Kal" could be of type "Age" with a literal value

I believe that this view is supported by the standard which defines a "topic
occurrence" as:
"Information that is specified as relevant to a given subject." (ISO
13250:2002 section 3.28)

The TMDM says (somewhat more cryptically):
"An occurrence is a representation of a relationship between a subject and
an information resource"

So if a topic occurrence can be used to convey information *about* a topic
(and is not just restricted to pointing to resources where the topic occurs
in the resource content), then perhaps it is more understandable that one
could claim that a topic name is a special form of information about a
topic. It is a label - it has the same relationship as a term does to a
concept in classification - and so you could deconstruct this to say "a
topic name is a label for a topic, the label of a topic is information about
the subject, and therefore a topic name is a topic occurrence"

Which brings me to Jan's point: IMO, the distinction between a label for a
topic and any other information about a topic is a fundamental part of basic
topic map semantics - to be able to label our concepts is such an important
aspect of dealing with them, that it is something which absolutely has to
have a consistent representation in the data model. 

Is a topic name information relating to the subject that the topic
represents? Yes, but its special information. Could we have just made a PSI
for a special occurrence type "Name"? Maybe, but issues abound around
variant names. However, the fact is that since inception the distinction
between topic name and topic occurrence has been in the standard and (IMO)
is a useful distinction to have that we would be foolish to remove.



> -----Original Message-----
> From: sc34wg3-admin@isotopicmaps.org [mailto:sc34wg3-
> admin@isotopicmaps.org] On Behalf Of Murray Altheim
> Sent: 29 September 2005 21:01
> To: sc34wg3@isotopicmaps.org
> Subject: Re: [sc34wg3] Topic names and occurrences in the TMDM
> Jan Algermissen wrote:
> > Murray,
> >
> > On Sep 29, 2005, at 6:56 PM, M.Altheim wrote:
> >>Hi,
> >>
> >>Just a question that came to mind whilst reading the section
> >>on Topic names in section 5.4 "Topic name items" of the TMDM:
> >>
> >>   [...]
> >>   Essentially, a base name is a specialized kind of occurrence.
> >>
> >>To my mind, names are not a specialized kind of occurrence at
> >>all -- names, occurrences and roles played in associations are
> >>the three kinds of topic characteristics. Topic names and
> >>occurrences are peers in that taxonomy; names are not a subclass
> >>of occurrence, which would imply that a name is an occurrence of
> >>the Topic, not a characteristic of it. That doesn't make sense
> >>to me.
> >>
> >>Clarification, please...
> >
> > Since TMDM suggested practice seems to be to express attributes of
> > subjects (age, height, eye-color, department number etc.) as
> > occurrences and since a name is just a special kind of attribute it
> > seemms logical to say that a name is a special kind of occurrence.
> No, it doesn't seem logical at all. Please don't include yet another
> term with "attribute". Let's simply call it what it's been called all
> along in Topic Maps: a Topic characteristic. We could step out to a
> more general knowledge representation language and use property, but
> since we're talking about the Topic Map model, we might stick with
> the language already extant.
> Now, if we have a Topic called "Person", then a Person's age, height,
> eye colour, etc. are just [typed] characteristics of Person, not
> occurrences of "Person". You and I are occurrences of Person, our
> names and eye colour are not. They are our characteristics.
> I literally today answered this same question regarding names in the
> Common Logic list. Names are characteristics (aka typed properties)
> of an entity. A property relation is a unary predicate (a 1-ary
> relation with the entity, i.e., in Topic Maps, a typed Association
> between a Topic and its characteristic).
> Unless the fundamental model of Topic Maps has changed since 2000 I
> don't follow the logic of this at all. The language is screwy.
> > Subject Indicators and Subject Address are then, BTW, also just
> > occurrences.
> Sorry, but that's also screwy. Anything to do with "subject" is
> a part of a Topic's identity-forming characteristics. Occurrences
> do not form identity except in the most remote sense of the class
> created extensionally by the members of its class, but I seriously
> doubt too many people would ever do that using Topic Occurrences,
> nor do I sense that this was the intent here.
> There was a time when everything was a Topic, and now we've reached
> a time when everything is a Topic Occurrence. *sigh*
> > I have allways wondered what the benefit of four constructs for
> > things that are just semantic variations of a basic functinality
> > could be....seems awkward to have to implement basically the same
> > thing four times....
> If you've read the graph theory description of Topic Maps you'd
> see that the simplifications we used were to make the model
> usable by humans. We could have decided not to have a Topic's
> characteristics be anything special, forcing them to be explicitly
> created via typed associations, but it's really a hell of a lot
> easier to use that tiny bit of syntactic sugar. I would on the
> contrary think it would have been *incredibly* awkward to have
> devolved the current model into a more basic graph structure --
> nobody would ever want to look at a Topic Map. This reminds me
> of the RDF renditions of XTM documents. Yes, it can be done, but
> no, it's not very useful.
> There are humans involved in this stuff. These shorthands are
> useful.
> > Oh well...does anybody care?
> I suppose one can take a nihilistic approach to ontology definition,
> but I don't. It really bugs me when language is turned upside down
> or distorted. Some of the knowing corruptions of legacy terminology
> in RDF are terrible, as if somebody didn't do any homework. We don't
> need to do the same in Topic Maps.
> Murray
> ......................................................................
> Murray Altheim                          http://www.altheim.com/murray/
> Strategic Services Development Manager
> The Open University Library and Learning Resources Centre
> The Open University, Milton Keynes, Bucks, MK7 6AA, UK               .
>         believe that everything is for
>         you until you discover
>         that you are for it
>                                "The Robin and the Worm" by Don Marquis.
>                                http://www.altheim.com/lit/robnworm.html
> _______________________________________________
> sc34wg3 mailing list
> sc34wg3@isotopicmaps.org
> http://www.isotopicmaps.org/mailman/listinfo/sc34wg3