[Top] [All Lists]

Re: [ontolog-forum] Topic Maps and OWL

To: "[ontolog-forum] " <ontolog-forum@xxxxxxxxxxxxxxxx>
From: srn@xxxxxxxxxxxxx
Date: 03 Jan 2004 15:52:00 -0500
Message-id: <m37k0892mn.fsf@xxxxxxxxxxxxx>
I think Mike Daconta's (hi Mike) analysis is possibly accurate, given
one view of what Topic Maps are, and of what the Topic Maps paradigm
is.  That view (the TAO view, as it's called by its popularizer, Steve
Pepper) implicitly holds that the Topic Maps paradigm is describable
as a single ontology whose primary features are certain relationship
types, including topic-occurrence, topic-name, topic-subjectIndicator,
class-instance, and superclass-subclass.    (01)

An alternative view of the Topic Maps paradigm is reflected in the
Topic Maps Reference Model now being considered for standardization in
ISO-land.  This alternative view is much more abstract.  According to
it, the objective of the topic maps paradigm -- let's call it the
"Topic Maps Objective" -- is to allow graphs (semantic networks) to be
predictably constructed and merged in such a way that there is a
single proxy for each reified subject -- one reifier per subject.  (In
the Reference Model, the terms "subject proxy", "topic", and "reifier"
are synonymous; they are typically implemented as objects.)    (02)

According to the proposed Topic Maps Reference Model, the means
whereby the subjects of proxies are discriminated, and, indeed, the
degree to which subjects are reified by proxies at all, are defined by
"Applications" of the Topic Maps paradigm.  "Applications" must
disclose these things.  (In the jargon of the Reference Model, an
"Application" is mostly an ontology -- primarily a set of relationship
types that, in the aggregate, constitute a universe of discourse -- a
world of sayable things.)  The Topic Maps Reference Model seeks to
establish the disclosure requirements that must be met by
"Applications", so that the degree to which the Topic Maps Objective
can be achieved within a given Application (within a given ontology,
and therefore within a given universe of discourse) is both explicit
and knowable.    (03)

So, for the Topic Maps Reference Model, there is no conflict between
OWL and Topic Maps.  Indeed, assuming that OWL (or an interpretation
of OWL) meets the disclosure requirements set forth in the Reference
Model, OWL (or at least an interpretation of OWL) can be regarded as
an Application of the paradigm, because the degree to which the Topic
Maps Objective can be achieved, when processing and merging multiple
OWL instances, is both explicit and knowable.  (Let me speak
specifically to Mike's concern here: under the Reference Model, there
is no limit to the robustness of the class modeling that can be done.
Applications define all of the kinds of statements that are statable.)    (04)

Subject proxy uniqueness -- another way of saying "the Topic Maps
Objective" -- is a demanding mistress.  She requires us to think:     (05)

* about every kind of subject,    (06)

* about how to discriminate every subject (be it a class, an instance,
  or both, or neither) from every other,    (07)

* about which distinctions are worth making and which should be
  blurred into more general or less precisely-defined subjects,    (08)

* about which subjects shall be reified and which shall not, and    (09)

* about other things, too.      (010)

And then she requires us to articulate all these design decisions
rigorously and explicitly.      (011)

I'm sorry to report that the ISO Topic Maps standard itself does not
yet fully articulate these decisions for either its HyTime-based
syntax or its XTM syntax.  I don't object to the TAO Application, at
least in principle; what bothers me about it is that its details
remain undisclosed.  The exact extent to which it supports the Topic
Maps Objective is, as yet, unknown.  This is why I say Mike Daconta's
view is "possibly accurate", given the incomplete state of the ISO
Topic Maps standard.  Indeed, under the present circumstances, I'm not
surprised that Mike worries about the "semantic validity" of Topic
Maps.  We can't expect predictable, deterministic information
interchange if we don't have a single rigorous, consistent and
comprehensively-articulated interpretation of the standard interchange
syntax.    (012)

Personally, I'd like to see two things:    (013)

(1) Consensus on and Reference-Model-conforming disclosure of the
    Application (the "TAO Application"?) that presumably undergirds
    the HyTM and XTM syntaxes.    (014)

(2) A Reference-Model-conforming disclosure of (if necessary, one or
    more interpretations of) OWL.  This worthy project would likely
    reveal much about OWL, and it might significantly advance the
    cause of the Semantic Web.  Jack Park says we should    (015)

      "...give due consideration to the one process requirement that
      makes topic maps what they are: you must merge objects which
      talk about the same subject.  I therefore don't think that
      anything out there today has superseded topic maps.  In the end,
      there are ways to construct ontologies such that they are, by
      default, topic maps, and nobody needs to know you did that."    (016)

    I agree with Jack.  I would only add, by way of clarification,
    that all that needs to happen is that the disclosure of the
    ontology meets certain requirements.  It doesn't have to advertise
    that it meets them, or claim that it meets them, it just has to
    meet them.  If and only if it meets them, the primary benefit of
    the Topic Maps paradigm -- proxy uniqueness for subjects, to some
    knowable extent -- can accrue.    (017)

Some references, FYI:    (018)

My talk at Extreme last year was an overview of the Reference Model;
you can find it at    (019)

http://www.mulberrytech.com/Extreme/Proceedings/html/2003/Newcomb01/EML2003Newcomb01.html    (020)

The latest draft of the Reference Model is available at:
  http://www.isotopicmaps.org/TMRM/TMRM-latest-clean.html    (021)

A web page on the Topic Maps Reference Model page, from which
the above is indirectly referenceable:
  http://www.isotopicmaps.org/tmrm    (022)

-- Steve    (023)

Steven R. Newcomb, Consultant
srn@xxxxxxxxxxxxx    (024)

Coolheads Consulting
http://www.coolheads.com    (025)

direct: +1 540 951 9773
main:   +1 540 951 9774
fax:    +1 540 951 9775    (026)

208 Highview Drive
Blacksburg, Virginia 24060 USA    (027)

*******************************************************    (028)

Peter Yim <peter.yim@xxxxxxxx> writes:    (029)

> This looks like a discussion a lot of us will benefit from.
> Both Topic Maps and OWL became "front runners" for good reasons. While
> there may be applications where both may be (almost equally) suitable,
> each of these approaches would be better suited for certain
> applications.
> What are those reasons (for their adoption so far)?
> What are the applications (best suited to each)?
> Where are the areas where they are pitched against one another (as
> well as other approaches)?
> Could those of you who are spending time on these issues throw some
> light on the matter, please.
> -ppy
> P.S. I am separating this into another thread, with a more appropriate
> subject (to ease future retrieval.)
> --
> ------- Original Message --------
> Subject: Re: [ontolog-forum] Re: [regrep-cc-review] What if? CCRIM =>
> Date: Thu, 1 Jan 2004 23:27:06 -0700
> From: Michael Daconta <mdaconta@xxxxxxx>
> To: [ontolog-forum]  <ontolog-forum@xxxxxxxxxxxxxxxx>
> Hi Everyone,
> The subject of Topic maps versus OWL is very important to work I am
> currently doing.  I am participating in a group that will be
> standardizing a DOD taxonomy framework and we will need to choose a
> representation format.  Clearly, topic maps and OWL are the two front
> runners.
> With regard to taxonomies, topic maps seem to have an advantage in
> simplicity, while OWL has the advantage in semantic validity.  Also,
> some of the semantic foundations are identical in that Topic maps do
> support class/subclass associations.  Unfortunately, the class modeling
> in topic maps is very weak.  Thus, one key judgement criteria is
> "how important is robust class modeling to effective classification?"
> Topic maps rely mostly on name-based classification (and merging) while
> OWL classes have robust properties, set theoretics and mapping.
> I plan on working through these details to support a recommendation in
> the next few weeks.  Comments, suggestions and insights are welcome.
> Best wishes,
>    - Mike
> -----------------------------
> Michael C. Daconta
> Chief Scientist, APG McDonald Bradley, Inc.
> www.daconta.net
> Jack Park wrote on 1/1/2004, 12:24 PM:
>   > Peter Yim wrote:
>   >
>   > >
>   > >>> Are topic maps out of the running?
>   > >>
>   > >>
>   > >>
>   > >> Some say you need both Topic Maps and OWL, though I cannot
>   > >> understand why. In my mind OWL supersedes Topic Maps..
>   > >
>   > Well, I can't let that one go uncommented.  There was an
>   > unfortunate "boxing match" between two Erics, one from topic
>   > maps and one from RDF.
>   > It was held as a humorous interlude at Extreme Markup 2000. When both
>   > Erics came on stage, they both sang praises of the other's "camp".
>   > Unfortunately, that's not what the press picked up on. Thus it was
> that
>   > RDF and XTM became "at odds" with each other.
>   >
>   > In the end, no matter what is said and done, OWL, RDF, whatever,
> and XTM
>   > or HyTM (sgml topic maps, the original ISO 13250 dtd) are
> serializations
>   > with which you can ship information around and have it arrive in a
>   > decypherable form at the other end of whatever wire is used. At
> the same
>   > time, each brings to the table some manner of underlying process model
>   > and semantics.  XTM, for instance, makes a minimalist ontological
>   > committment to objects necessary to capture topics, which are known as
>   > "the place you go to find out everything that is knowable about a
>   > particular subject" and a subject is "anything you can talk
> about."  The
>   > topic maps underlying process model simply dicates that if two topics
>   > are about the same subject, they must be merged.  AFIK, OWL doesn't
>   > require such processing. I would therefore respectfully submit
> that, if
>   > you happen to need the organizational power that comes with topic
> maps,
>   > no matter how you construct them (yup, they have been written in OWL),
>   > then you must give due consideration to the one process
> requirement that
>   > makes topic maps what they are: you must merge objects which talk
> about
>   > the same subject. I therefore don't think that anything out there
> today
>   > has superseded topic maps.
>   >
>   > In the end, there are ways to construct ontologies such that they are,
>   > by default, topic maps, and nobody needs to know you did that.
>   >
>   > 2004 is shaping up quite nicely already!
>   > Cheers
>   > Jack
>   >
> _________________________________________________________________
> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
> Subscribe/Unsubscribe/Config: 
> Shared Files: http://ontolog.cim3.net/file/
> Community Wiki: http://ontolog.cim3.net/wiki/ To Post:
> mailto:ontolog-forum@xxxxxxxxxxxxxxxx
>      (030)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (031)

<Prev in Thread] Current Thread [Next in Thread>