[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

Re: Topic Maps - current state of the art?

• From: Steve Newcomb <srn@coolheads.com>
• To: xml-dev@lists.xml.org
• Date: Sun, 20 Oct 2013 00:42:51 -0400

On 10/19/2013 08:14 PM, Stephen Cameron wrote:
> Topic Maps are something I have just discovered!
> ...
> Thanks for any insights you can offer.

I see that I have written a lot below.  Sorry, Stephen.  I guess you
have pushed one of my buttons.  And I'm not sure I answered your
question about the "state of the art" of Topic Maps.  Instead, I tried
to answer a shorter, even more interesting question: "What is the art of
topic mapping?"

I wrote most of Chapter 2 of the book you mention.  That chapter is
still offered free on the web by Addison-Wesley at
http://www.aw-bc.com/samplechapter/0201749602.pdf.  With Michel
Biezunski, I also drafted the original Topic Maps standard; a remarkably
large portion of that draft, for better or worse, survives today as ISO
13250.  We did what we could with the time, skills, and insights that we
had at the time.

I've written more about topic mapping, much of which appears at
http://www.coolheads.com/publications.htm.  My personal favorite,
co-written with Patrick Durusau, can be found at
http://www.coolheads.com/SRNPUBS/EML2005NewcombDurusau/EML2005Newcomb01.html.
 It's a concrete and homely demonstration of what we were talking about
all along.

Michel, Vicky Newcomb and I maintain the IRS (U.S. Internal Revenue
Service) "Tax Map" product on an at-least-weekly basis, and more often
during tax season, as we have done for over a decade.  A public version
of IRS Tax Map can be found at http://taxmap.ntis.gov, but only on days
when the U.S. government is open for business.  Comments about it should
be addressed to topicmaps@irs.gov.

From my personal perspective, which is not universally shared in the
small but worldwide group that identifies itself as the Topic Maps
Community, the whole idea of topic mapping is far more abstract than any
particular syntax or any particular semantic model.

Just a little history here: The idea of topic maps grew out of, and was
explicitly based on, the HyTime (ISO 10744) ideas of architectural
forms, groves, and grove property sets.  Without an understanding of
those ideas, and/or without incentive to meet the public-spirited
requirements that those ideas were designed to meet, the basic idea of
Topic Maps was confounded with that of its original formal model: the
Topic Maps meta-DTD.  Soon the meta-DTD was universally understood as an
ordinary DTD, which, in turn, was misunderstood as implying or
demanding a model tantamount to a specific UML diagram.  That model, in
turn, was implemented as the Ontopia suite of tools and products, and in
a number of open-source projects, as well.  Talented, hardworking people
did a lot of fine work on that, although I always thought they were
missing the point.  And so "Topic Maps" turned out to be the name of
something almost the opposite of what I hoped it would be; it became the
name of a standard for a particular model, instead of a way of thinking
about, and approaching the problem of providing for, enhanced human
communication.  ("Global knowledge interchange" was once the catchphrase
I favored for the latter goal -- hence the title of Chapter 2 in Jack
Park's and Sam Hunting's book.)

By now, my earlier passions have been slightly moderated by practical
experience and advancing age.  I now think the *original* idea of topic
mapping boils down to an esthetic -- a Muse that demands a certain kind
of intellectual discipline and diligence in exchange for her favors.
The Muse's favors are real enough; she really does offer improved
prospects for human understanding and global prosperity.  However, the
Muse does not favor those who worship other gods, such as the
private-sector god of next quarter's share price, or the public-sector
god of one's career prospects in the coming budget year.  Topic Maps
just don't seem to work that way.

(It remains a puzzle exactly why the Muse is so jealous in that way.
Isn't it more natural and rational to work toward a state of affairs in
which everybody is rich, rather than to be more focused on being richer
than others?  This ancient question seems more urgent, in many ways and
for many reasons, every time I watch the PBS Newshour.  But I digress.)

What does the Muse of Topic Maps demand?

(1) Topic mappers must deeply understand that a topic map is neither
more nor less than an editorial product.  A good topic map is about as
likely to occur by accident as a good book, movie, or any other work of
art.  For better or worse, a topic map will always reflect its
maintainer's intentions, skills, attention to detail, and the practical
constraints under which the work of maintaining it is done.  A topic map
cannot be generated by a computer, except in the same sense that a
symphony or book can be computer-generated.  Computers may be interested
in such products, but human beings will not find them compelling.

(2) A topic map is a representation of a space in which subjects of
conversation have unique addresses.  The residence of a subject of
conversation -- the thing that bears an address -- is called a "topic",
a term which in topic-mapping-land is a term of art.

A topic always "reifies" -- thing-ifies -- exactly one subject of
conversation.  A subject of conversation is anything anyone honestly
thinks he or she is actually discussing.  A single subject can live in
multiple residences simultaneously, i.e., it can be reified by any
number of "topics".  The Muse does not demand any particular reification
technology; a topic *can* be implemented as an object, and it might
*not* be implemented that way, too.  The only requirement of a topic is
that it must *explicitly* represent exactly one subject -- whatever the
word "represent" may mean in some context.  By "explicit" I mean that
the identity of its subject must be explicitly and uniquely represented.
 If two topics bear the same subject address, the same subject actually
does live there, too.  Also, a topic can bear any number of addresses
greater than or equal to 1, as long as the topic's creator really means
that one and the same unique subject has all those addresses.

(3) Everything else the Muse demands is a matter of the characteristics
of *useful* topic maps.  The Muse asks, in exchange for her favors, that
the author(s) of a topic map must care about:

(a) The characteristics of the address space.  Two people who have found
reifiers for a single given subject of conversation should also be able
to discover that they are in the same residence, or at least in two of
the many residences at which that same subject resides.  There are many
issues here, for example: Is the address derived (or is it derivable)
from the connections between the topic and other topics?  Are the
connections also addressable (do they have addresses, i.e., are they
reified)?  If so, where does the reification of connections end, i.e.,
what connections remain unreified?  (Some must.)  Is there an outside
identity-authority?  If so, how does it work?  If not, how is the need
for its services avoided or finessed?

More substantively: Does the address space fairly reflect how those who
work in that space -- in that universe of discourse -- actually think
and communicate with one another?  Does it support semantic evolution
and drift, without unnecessarily compromising the usefulness of earlier
work in the ancestral space(s)?  Does it support *collaboration* on the
evolution of the address space?  In general, can everybody tell, more or
less, what everybody else is now talking about, or has been talking about?

(c) The characteristics of the reification technology.  Does the
technology usefully scale to the size of the address space?  Can it
perform adequately for the purpose?  How is it interchanged?  How can
people collaborate?  How can independently-maintained topic maps,
expressed in terms of independently-maintained address spaces, be
exploited in each other's contexts?  How will a given address space
support (or hinder) the efforts of those who seek to build bridges
between address spaces -- between universes of discourse?  What primary,
secondary, and tertiary business models are supportable?

(d) In view of all of the above considerations, are all of the following
perspectives compellingly reasonable facsimiles of some unified view of
the realities about which some address space purportedly enables
communication:

   i. From the perspective of a given topic (or, more accurately, from
the perspective of a single given subject): how does any other topic
(read: subject) appear or not appear?  What do I need to know in order
to move between any two topics?  How well does the analogy between the
subject being reified and its topic stand up to stress and scrutiny?

   ii. From the perspective of a topic/subject, how does it see itself
as a part of the whole space?  What's missing, finessed, or ambiguous?
Are the lacunae, finesses, or ambiguities compellingly (i.e., usefully)
reflective of the universe of discourse being represented?  (Note: This
is where I think the topic mapping and the ontology-oriented communities
enjoy a bit of dialectic tension with one another.  The Topic Maps Muse
does not require scientific consistency, nor does she necessarily expect
that conclusions will follow logically as they might from some rigorous
and well-specified ontology.  Our Muse simply demands that a topic map
accurately reflect how some individual or community actually thinks.
Remember: a topic map is an editorial product.  At its best, the art of
topic mapping enables us to sense the real existence of other
individuals who are like us, who can share their perspectives with us,
and who are as unique as we are.

  iii. From the perspective of the whole address space/universe of
discourse, how does it differ from other such spaces, and what does it
have in common with them?  What class of address space is it?  And what
are the classes of address spaces?

   iv. Again from the perspective of the whole address space, how does
it see the topics/subjects it, uh, may contain?  What introspection
facilities does it provide?  Does it include topics about what
topic-ness is?  Does each address represent *two* subjects (one an
actual subject and the other the subject which is the topic itself
considered as a topic of conversation)?  Which metadata are unreified as
topics?  (Personally, I think "metadata" is a euphemism for
underprivileged data.)

• References:
  • Topic Maps - current state of the art?
    • From: Stephen Cameron <steve.cameron.62@gmail.com>