Consistency and Formal Model Re: [topicmapmail] Can a resourceRef be a topic

[Bernard Vatant]

Steve, Peter

Peter's TM *implicit model* misinterpretation and Steve's remarks about it - that I second
completely - show once again that we definitely need a formal model that clearly forbid
that kind of semantic inconsistencies, like using the same topic reference for "topic
type" and "scope", or using a topic that represent an individual object as a class, and
the like ...

I jump on the occasion to announce that Mondeca graph university team has submitted a few
days ago a paper for ISWC 2002 conference, proposing a mathematical model for topic maps -
based on hypergrahs, completely written in formal set theory. This model introduces the
notion of semantic layers in topic maps, and proposes a consistency constraint that should
forbid that kind of semantic mismatch.

The paper has been "semi-published", meaning that a few brains in the TM community has
been contacted last week for review during the final making of the paper. This was kept
inside a limited group to avoid a debate overload that would have certainly led the
authoring team to be unable to meet the submission deadline ... please those who were not
asked don't feel like outcasts, you'll all have soon a chance to fight with that stuff :))
The feedback we got so far is that there is a lot of communication and explanation work to
do to show how this formal model fits the needs. And since I've accepted the challenge to
be the bridge between the mathematicians and the TM community, I will try to deliver ASAP
an on-line presentation of that model, readable for non-too-much-mathematicians. A
challenge indeed ...

To those who would like to have a look at the original paper anyway "as is", with much
maths inside, but also some prose and graphics around (postscript format) I can forward
it - with of course all usual appending copyright and confidentiality conditions -

Bernard

----- Message d'origine -----
De : "Steve Pepper" <pepper@ontopia.net>
À : "Topic Map Mail" <topicmapmail@infoloom.com>
Envoyé : mercredi 20 février 2002 10:40
Objet : Re: [topicmapmail] Can a resourceRef be a topic


> At 10:04 19/02/02 +0000, Sparkes, Peter wrote:
> >In a topicMap, if individual topics have a Parent/Child relationship, can
> >the child relationship be expressed using:
> >
> >         <topic id="BA">
> >           <occurrence>
> >             <scope><topicRef xlink:href="child"/></scope>
> >             <resourceRef xlink:href="BA1"/>
> >           </occurrence>
> >         </topic>
> >         where the resourceRef points to a child topic with an ID of "BA1"
>
> Although this is syntactically correct, it flouts the topic map model. You
> should not use an occurrence to express a relationship between two topics,
> *unless* one of the topics happens to be an addressable information
> resource that is in some way pertinent to the other topic (which does not
> seem to be the case here).
>
> To express parent-child relationships, use associations, thus:
>
>    <!-- assume topics for "parent-child", "parent", and "child" -->
>
>    <topic id="BA"/>
>
>    <topic id="BA1"/>
>
>    <association>
>      <instanceOf><topicRef xlink:href="#parent-child"/></instanceOf>
>      <member>
>        <roleSpec><topicRef xlink:href="#parent"/></roleSpec>
>        <topicRef xlink:href="#BA"/>
>      </member>
>      <member>
>        <roleSpec><topicRef xlink:href="#child"/></roleSpec>
>        <topicRef xlink:href="#BA1"/>
>      </member>
>    </association>
>
> Or, in LTM syntax:
>
>    parent-child( [BA] : parent, [BA1] : child )
>
> >The parent relationship can be expressed using:
> >
> >         <topic id="BA">
> >           <instanceOf>
> >             <topicRef xlink:href="B"/>
> >           </instanceOf>
> >         </topic>
>
> This is also incorrect: It expresses a class-instance relationship, not a
> parent-child relationship. To see this, replace "BA" with "Peter" and "B"
> with your father. Is your father a class? No. Are you an instance of your
> father? No. Once again, an association is the way to go:
>
>    parent-child( [B] : parent, [BA] : child )
>
> >I want to do this so that every topic contains the xlink to its parent and
> >children.
>
> If I understand correctly, you have a *syntactic* requirement for not using
> association elements. In that case, the XTM syntax does not meet your
> requirements and you need to devise your own syntax to do so. This need not
> be a problem, since the XTM syntax is only intended for *interchange*.
>
> However, you should be aware that you will have a redundancy problem if you
> want every topic to contain links to both its parents and its children. For
> example, the relationship between [B] and [BA] will be expressed twice:
>
> * once from [B] ("my child is [BA]"), and
> * once from [BA] ("my parent is [B]").
>
> This is precisely the reason why association elements are not subelements
> of topic elements in the XTM syntax: An association usually involves more
> than one topic, and doesn't "belong" to any one of those topics in particular.
>
> If your syntactic requirement is important enough and you think you can
> handle the redundancy problem, devise your own syntax and generate XTM
> automatically whenever you want to interchange your topic map. Otherwise,
> find a satisfactory way of eliminating the syntactic requirement. (Using a
> conforming topic map engine would be the best way to go.)
>
> Best regards,
>
> Steve
>
> --
> Steve Pepper, Chief Executive Officer <pepper@ontopia.net>
> Convenor, ISO/IEC JTC1/SC34/WG3  Editor, XTM (XML Topic Maps)
> Ontopia AS, Waldemar Thranes gt. 98, N-0175 Oslo, Norway.
> http://www.ontopia.net/ phone: +47-23233080 GSM: +47-90827246
>
> _______________________________________________
> topicmapmail mailing list
> topicmapmail@infoloom.com
> http://www.infoloom.com/mailman/listinfo/topicmapmail
>