"Florian G. Haas" wrote:
[...]
> Does anyone agree with me that it's *desirable* to use a <topicRef> for this
> purpose where you can, and resort to a <subjectIndicatorRef> if you must? Or
> am I completely out to lunch?

Yes, the design was created to allow either, but as you state, it's desirable
to use an explicit topic reference when it is to a <topic> element. This was
what we had in mind when we wrote the DTD.

> That being said, I'd like to voice a personal opinion about the discussion
> and approval process of country.xtm and language.xtm. Firstly, it's good to
> see that there's some progress. Secondly, I'm not personally acquainted with
> either of the two gladiators involved, and I firmly believe that whatever
> selection is eventually made, it had best not be influenced by any
> particular person's authorship. IMHO, what should be selected should be the
> best available. And then of course there's the option of calmly sitting down
> and comparing the two versions, and maybe creating a third one that combines
> the advantages of both...

All I've been asking to do is wait until we have a mediation process to
deal with the disagreements. It's silly for us to argue for weeks over
these issues with no way to really resolve anything. We're two weeks
away from having a process, and then I'm quite happy to deal with these
things then.

Murray

...........................................................................
Murray Altheim                         <mailto:murray.altheim&#x40;sun.com>
XML Technology Center
Sun Microsystems, Inc., MS MPK17-102, 1601 Willow Rd., Menlo Park, CA 94025

       In the evening
       The rice leaves in the garden
       Rustle in the autumn wind
       That blows through my reed hut.  -- Minamoto no Tsunenobu

Murray Altheim and I debated at length on this list how to interpret
relative URI references and the URIs of elements in the presence of
xml:base attributes. I believe I have now, with a little help, been
able to find out how this is supposed to work.

The answer satisfies me, but I believe that most XTM implementors will
be surprised by the answer. I posting my understanding of this issue
here, both for the record, and also so that anyone who thinks I am
wrong can correct me.

Here is an example, to make things a bit clearer:

(in a document residing at "http://www.quux.com"):

   <topicMap xml:base="http://foo.com/">

     <topic id="bar">
       <instanceOf>
         <topicRef xlink:href="#baz"/>
       </instanceOf>
       ...
     </topic>

     ...

   </topicMap>


The main questions were:

   - what is the URI of the "bar" topic?

   - what URI is "#baz" referring to?


It turns that the answer to the first question is not defined
anywhere, but that the only sensible answer is
"http://www.quux.com/#bar". That is, the URI is not affected by the
changed xml:base inside the document.

The reason this is so has to do with the answer to the second
question. URIs found in a document are always in a situation where
that document has a document URI, and this may have been modified
inside the document, so that inside the document a different base URI
is in effect.

It turns out that there are not two kinds of URI references, as most
people think (until recently this group included me), but three:

  - absolute URI references.  These are, of course, independent of the
    base URI of the context they appear in.

  - relative URI references.  These are resolved relative to the base
    URI of the context.

  - same-document references.  These are of the form "#xxx" and are
    resolved not relative to the base URI, but relative to the document
    URI.

So because "#baz" is a same-document reference it resolves to

   http://www.quux.com/#baz

Had it been "baz.html" it would have resolved to

   http://foo.com/baz.html


The reasons I believe the above is correct can be found in:

<URL:
http://lists.w3.org/Archives/Public/www-xml-linking-comments/2001AprJun/0158.htm\
l >

and in RFC 2396, section 4.2.


In the particular discussion Murray and I had, he was right with
respect to what the subject indicators of the topics in the
language.xtm and country.xtm topics were, and I was wrong.

I am not sure that his explanation of the general rules for this
situation was right, but I didn't fully understand it, so it may have
been.

--Lars M.

Hello,

Just a quick note before anyone gets too excited (pro or con) about
adopting "official" positions at the topic map meeting after the Extreme
Markup conference in Montreal.

The SBL recently joined OASIS, in part so I could continue to work in
the new group, so I am not yet familiar with all the ins and outs of
OASIS procedure. Assuming that a new group is formed, IMHO, a very good
idea, I don't think we will be in a position (as an OASIS group) under
OASIS rules to adopt anything at the Montreal meeting. I may be
mis-reading the OASIS rules completely, so long time members of OASIS
please correct the foregoing if I am in error.

I certainly think we should have focused and meaningful discussion on
all the pending proposals which may well form the basis for later action
of the group. Let's not cloud the discussion with expectations of
"official" approval just yet. I have found the recent discussion on PSIs
and other technical work quite interesting (although due to server
migrations and a host of other activities I have not been able to chime
in) and in line with how I see a technical working group moving forward.
Not everyone agrees but the discussion has lead to greater examples and
more detailed arguments, which I find helpful in exploring these issues.

Looking forward to seeing everyone in Montreal!

Patrick

--
Patrick Durusau
Director of Research and Development
Society of Biblical Literature
pdurusau@...

Hello again,

while I was going through the spec examples in research for my previous
post, I noticed a mishap in paras F.6.3 and F.6.4:

<association id="a34">
	 <member>
		 <roleSpec xlink:href="#brother" />
		 <topicRef xlink:href="#gdm" />
      	 <topicRef xlink:href="#ctb" />
       </member>
	 <!-- ... -->
</association>

Something's just not quite right about the <roleSpec>... :-)
That error occurs about 8 times in those two paragraphs. Editors, please fix
this.

tia
-- Florian

----------------------------------------------
Florian G. Haas
FHS Informationsberufe, Eisenstadt, Austria
http://info.fh-eisenstadt.ac.at/~haasf/

* Murray Altheim
|
| No, this was the crucial point. They did *not* standardize the
| languages and countries of the world, they standardized a list of
| their *names* as well as a set of codes as proxies for those
| names. Look at the name of each standard. Both ISO and IETF always
| call them "codes for the representation of languages", "Tags for the
| identification of languages," etc.

So, in other words, what the codes do is to represent or identify
languages.

Let's be a bit more specific here, so that it is easier to make sure
we are on the same page. What, in your opinion, is the subject of this
topic?

   [german
    @"http://www.topicmaps.org/xtm/1.0/language.xtm#de"]

(For those who don't know it: I'm using LTM notation here to create a
topic with the subject indentifier that is the URI above.)

In your opinion, is the subject of this topic the langauge code 'de'
or is the language German?

If it is the language code 'de' then I wonder what the PSI can
actually be used for. Can you make some examples of statements made
using the topic that would seem sensible to you?

Also, if, in the same LTM file, I were to put in the association

   spoken-in([german], [germany])

what, in your opinion, would be the assertion made by that
association?

| It's fallacious to do otherwise, and the current language and
| country topic maps do exactly the same thing, by conscious design.

I know it was by conscious design, but like Kal I am having great
problems understanding the rationale for that design. Please help me
out here and try to explain how you think your approach is the
sensible way to do it.

* Lars Marius Garshol
|
| Furthermore, the current version of language.xtm says that the
| language code 'de' has two names:
|
|   'de'
|   'German'
|
| I think it is reasonable to claim that the name of the language code
| is 'de', but to claim that the code has the name 'German' is absurd.
| There is a language, known by the names 'German', 'Deutsch', 'tysk',
| and many others, as well as by the code 'de', but the code itself has
| no name but itself.

* Murray Altheim
|
| This is either a red herring or you've not read the topic map very well.
| The 'de' string is scoped by being an alpha-2 code. The 'German' string
| is scoped by "English" meaning it's an identifier for anything German
| when you're speaking English. Let's be clear here.

The 'de' string is a characteristic of a topic, which has some
subject. If the subject is the language 'German', then that is an
instance of the topic 'language', and not 'language-code', right?

To me 'de-the-language-code' and 'german-the-language' are two
different subjects, which must be represented by different topics.  It
follows that the characteristics of the one cannot be assigned to the
other.

In other words, the string 'German' is the name of a language, not of
a language code. The name of the language code is 'de', and the code
can have no other name.

This is how I think about this. Obviously, your thinking differs, but
I haven't yet comprehended how.

* Lars Marius Garshol
|
| In short, I can't see how these topic maps can be of any use to anyone
| unless the topics in them are instances of 'country' and 'language',
| respectively. Surely that must be the point of the whole exercise?

* Murray Altheim
|
| Your argument holds no water; ISO and IETF have been doing precisely
| the same thing as these topic maps for many years.

Well, have they? Here is a quote from RFC 2616 (HTTP/1.1), section 14.4:

   For example,

     Accept-Language: da, en-gb;q=0.8, en;q=0.7

   would mean: "I prefer Danish, but will accept British English and
   other types of English."

Now, if this doesn't mean that the IETF uses the language code 'da' to
mean the language Danish I don't know what it means.

--Lars M.

Hello all,

now for something completely different -- and for the purpose of getting
away from personal animosities, wounded egos, and the like -- I'd like to
raise a point about the applicational or implementational suitability of
these two topic maps. Maybe this is not a big deal, but I'd be happy to read
some of your thoughts.


Lars Marius frequently uses <subjectIndicatorRef>'s whose xlink:href
actually resolve to a <topic> element and *could* therefore also be replaced
by <topicRef>'s (at least the spec says so, para 3.2.1). Now I'm fully aware
that that's not a requirement and that without question, the
<subjectIndicatorRef> element is syntactically fine.

In my humble opinion, though, using a <subjectIndicatorRef> when when one
could also use a <topicRef> is a questionable practice. In a nutshell, how
is an application supposed to be able to discern what actually points to a
<topic> (and can therefore be resolved, a <baseName> in the user's favorite
<scope> can be selected, and a #display variant displayed, for example), and
what doesn't (and in fact, may not even point to an XML node)?

Does anyone agree with me that it's *desirable* to use a <topicRef> for this
purpose where you can, and resort to a <subjectIndicatorRef> if you must? Or
am I completely out to lunch?


That being said, I'd like to voice a personal opinion about the discussion
and approval process of country.xtm and language.xtm. Firstly, it's good to
see that there's some progress. Secondly, I'm not personally acquainted with
either of the two gladiators involved, and I firmly believe that whatever
selection is eventually made, it had best not be influenced by any
particular person's authorship. IMHO, what should be selected should be the
best available. And then of course there's the option of calmly sitting down
and comparing the two versions, and maybe creating a third one that combines
the advantages of both...

Ceterum censeo: Now I'm happy to accept any flames. :-)

Take care
-- Florian

----------------------------------------------
Florian G. Haas
FHS Informationsberufe, Eisenstadt, Austria
http://info.fh-eisenstadt.ac.at/haasf/

* Murray Altheim
|
| [...] the group must now decide between the ones I've authored and
| these new ones.

Not so. It would be perfectly reasonable for the group to pick the
best aspects of each topic map to create the best possible end
product.

| Since my revised language and country topic maps do everything that
| Lars Marius' does (as far as I can see),

Not everything. I would guess that they don't actually state that
German is a language, and that they do not provide a PSI for the
concepts of countries and languages.

| [...] doesn't do things I disagree with (such as remove the
| necessary 'xml:base' attribute

Could you explain why it is necessary?

| [...] or make the XTM DTD URI absolute, which will just cause
| parsing headaches),

Could you explain how that causes parsing headaches worse than a
system identifier pointing to a non-existent local file? I'm not dead
set on the one approach or the other, but I would like to know the
reasoning behind yours before agreeing or disagreeing with it.

--Lars M.

Paul & John:

John's blazing entrance is perhaps one of the most exciting things to come
of Paul's assistance in widening the circle of our conversations. John is in
my view one of the most dedicated, prolific, and scholarly writers on
knowledge representation that I know. As some of you may know, at least 1/3
of the Ontology paper Rob and I have been writing and passing back and forth
is devoted to our expansions of John's basic ontological categories. This
expanded set is being built into our new Mark 3 Knowledge Mapping Tools.

John's publication last year of "Knowledge Representation -- Logical,
Philosophical, and Computational Foundations", John F. Sowa, Brooks/Cole,
2000. Is an absolute must buy for this field. John and I had spent some time
together in 1994 at his Conceptual Structures Conference in Maryland. He was
just starting the deep study of the historical roots of ontology evidenced
in this Brooks/Cole book. I had since lost touch with him and his
reemergence "to me" is most opportune.

John has indicated he will review the ontology paper (for now called,
Ontology 5) in the next few weeks. Some of you may have gotten an earlier
version Ontology 3 by e-mail.

This brings to mind that I have no idea what sort of crowd and interest
areas this beadgame has put me in. Perhaps Paul or the "moderator", if thats
the word, will venture some characterization, off line or on.

I can characterize the group I brought with me (the last four lines of Cc...
Names.) Most represent an exceptional group of engineers, project managers,
and pioneering software & hardware developers. A few are writers, charged
with articulating knowledge management and knowledge engineering theory and
practice to CEOs, CKOs, ... A few are long term sponsors and program
managers who must turn knowledge engineering potential into specific
national assets (laboratories, ships, airplanes, rockets, ..) or clearly
articulate their services and benefits for civil servants, admirals,
generals, and the President.

Not a few have been using these technologies for a decade or more and have
their own ideas on what part of current theory and technology works and what
part is just so much smoke. They are curious, exceptionally bright, but
their customers are very concrete. If they try to sell theory they will be
thrown out and never let back in. If these ideas are not being tested in the
real world, then go on and teach them in schools, but you are better off
getting some real consulting experience. Drop names , argue from authority
and supposed rigor, or community values and these people (me included) blow
you off. Anything worth telling their clients must be clearly pictured,
plainly spoken, and filled with real world examples.

******** Paul, does this give you some fix on their and my agendas?

My job has been to live both in experimental science, theory, engineering,
philosophy, teaching, project management, business, corporate formation,
yada, yada , yada. for 30 + years. I love every part of it. We are at one of
the greatest places and times in history. I get to put together theories of
knowledge one year and turn them into tools and nationally important
products and services the next. These are my friends and colleagues, next
year I will give them yet another generation of tools and together we will
begin project building and tool testing again.

*********

Let me take up the point where John came in and give you both the word
associations that will let you and our audience see that what we are talking
about has been "core" philosophy for over 2000 years. Seven years ago I gave
John the diagram I have enclosed below as a PDF file (Knowledge Divisions).
I had constructed this diagram a week or two earlier for a panel discussion
on Art, Philosophy, and Metaphor at the Massachusetts College of Art.

It is intended to provide that clear and global picture of what it is we are
talking about when we talk of knowledge. Now John, mindful of the positivist
(Vienna Circle) objections to the word "metaphysics" uses "abstract"
systematically in his writing. But basically this picture captures the great
realms of philosophy: (1) physics -- all description of the world of things
and activities we can sense and measure directly and (2) metaphysics (above
physics) -- the world of thoughts, ideas, theories, forms, etc. out of which
we construct our internal (mental) models and constraints when viewing or
acting in the world.

Abstractly, the triangular shapes on both sides are abstraction hierarchies
(placed on their sides). All of physical reality is on the right, its
abstraction is in aggregation where every sensible particle or quantum lies
nearest the center right and everything else that is physical is some larger
object or energy aggregated and then buried in still larger physical
conglomerates. The apex at extreme right might be called the "universe"
unless we want to leave room for multiple universes and realities in some
still to be defined physical sense.

At the very center, we have representations -- the language, codes, and art
by which we try to turn all knowledge into something permanent and enduring.
We hijack some arbitrary physical phenomena and teach each other from birth
that what we are sensing of its properties represents some particular
message and meaning.

The metaphysical side is the the most difficult to talk about because we
cannot see thought so clearly as object. Some hold hope that brain scans
might somehow do this, but most experiments suggest "ideas" are like
holograms and distributed broadly and redundantly, that what we see is just
the retrieval process. All that speculation will work its way out, the point
is that it is still exceedingly difficult. Whatever the fate of tying brain
physiology to thought labeling, no philosopher propose that we wait for a
completely physical view to somehow predominate.

We have instead in metaphysics hierarchy of ideas in this picture, that
moving to the left are progressively more global in their coverage. The
arrangement process is one of abstraction, dealing progressively with terms
or concepts that systematically ignore greater and greater differences in
order to concentrate still more on some more brightly shining and exemplary
ideal or similarity.

Most philosophers accept that the "least abstraction" is the plural
distinction. One real thing is physical. The collective term crosses the
line and is abstract. Two things implies abstracting and ignoring some
distinction that gives separate identity. The exclusion principal in
physics, produces different physical laws for truly identical objects. This
suggests there is something still more fundamental to be learned from this
issue of plurality and difference.

At the other metaphysical limit, the apex of this abstraction is "all
knowing" to some, here theologies claim their own naming conventions.

For real world knowledge bases, the terms model or archetype or less
rigorously "blueprint" are often used to designate the "least abstract"
abstraction. On the other, physical side is the "exemplar" -- the real
object or action or property that best represents the abstract idea. This
crossing relation between physical and metaphysical examples is most
critical to knowledge editing. This is all just today's way of talking about
the "universal" / "particular" distincts that Plato and Aristotle debated
over 2000 years ago. Every one since has used these to construct their own
philosophical definitions and examples. Our (Concept Sub-Codes) denote
"universal models" and (Instance Sub-Codes) designate "particular real
examples."

"Swing-wing fighter" is a abstract concept, "F-14" is another abstract
concept. Moving down some purely metaphyical taxonomy. Both are coarse
grained distinctions. Our customers would want something less abstract. F-14
is a "Design". They might want a "Aircraft Model" designation, like F-14A.
Our editors (several of those copied above) work regularity to push abstract
levels ever deeper to "procurement block" or "production lot", etc, etc.
Then cross over into real exemplars and examples (tailnumbers).

Their greatest problems come in integrating knowledge from different sources
and from authors who switch granularity to hide poor research and
specificity, e.g.
authors who make broad assertions about a given design, when what they say
is true only for certain models. These are typical of the semantic problems
our tools and people must deal with in integrating across layers. Our codes
let us follow all these ups and downs and stay unique and uncompomised by
language differences.

To say the universal / particular relationship spans every form of knowledge
is no exaggeration or stranger to philosophy. Experimentally, for 15 years
it has given us no cause to look for something else.

This conversation can go elsewhere, finding ever increasing use of
ontology-based coding. These parts of theory are not theory nor speculation
any more. For engineers and academics interested in them, I suggest you look
at George Lakoff's book,  "Women, Fire, and Dangerous Things" it is a most
complete analysis of the model - sub-model - exemplar clusters and the
impact of physical and psychological factors on category construction
criteria, in primitive and in modern societies and languages.

I was filling in for him at the conference when I drew up the diagram in
front of you.

For still more practical examples look at ICD-9-CM and CPT Code linkages.
Every technical book store has a shelf or bookcase filled with medical
concept coding books. These are built, not by academics, but by publishers,
engineers, and systems analysts looking for a way so that doctors can write
down codes in patient records that will then let some $5/hr clerk match this
patient up with insurance company rules, guaranteeing that their experts
will accept the treatment plan and pay the bill. This is all too important
to trust to language or to all the educational differences that exist within
this chain of communications and work culture.

We talk often that the goal in knowledge engineering is to put beneath the
human practioner a floor of externally encoded knowledge, that the measure
of that knowledge is the success a novice might have using it as compared to
an acknowledged expert when faced with a spectrum of problems. This is not a
hypothetical, it is already a reality where the para-professional
(para-medical, para-legal, etc) are replacing the supposedly more
experienced professional. My closest friends in medicine, themselves
experienced doctors, say the results are in and the para-professionals do
the better job.

For them, the future of the high professionals in medicine is in making new
knowledge, not in dispensing what is already known. This brings us to a line
that Paul feels we must never cross.

Given the very practical nature of this audience and their great infuence on
computing, is this the pulpit you want to speak from to make your case that
these goals or ambitions are, at best, bad policy or inevitably doomed to
failure because intelligence is and must always be biological. Am I
mis-stating your position?

Is their an unbridgeable distance between a human novice using the
computer's encoded knowledge store and the machine functioning on its own.
Where is this line and why?

Nuf said.

Dick






-----Original Message-----
From: John F. Sowa [mailto:sowa@...]
Sent: Wednesday, July 25, 2001 8:23 AM
To: Paul Stephen Prueitt
Cc: Brian (Bo) Newman; Richard Ballard; Protege-Discussion; Ray
Fergerson; Peter Kugler; Andrius Kulikauskas; Complexity Group; Ivan A.
Uemlianin; Jeff Woodword; Steven R. Newcomb; Xtm-Wg; W. M. Jaworski;
Int_group; Dave Lehman; Rob Smith; beadgame; Jon Awbrey; Adriaan Jooste;
Angelos Kolokouris; Don Cates; Fred van Leeuwen; Frederick G. Schobert
Jr.; Gordon Runner; Jake Lefman; Jerry M. Haber; John Kerr; Joseph
Williamson; Kurt W. Conrad; Malcom P Taylor Jr.; Mark Turner; Mary E.
Bjustrom; Nami Ataee; Richard (Dick) Knudson; Steve Barth; Terrill M.
Moore; Tom Embrogno; William Capps; 'Com-Prac; Steve Pepper
Subject: Re: an introduction to the BCNGroup beadgames


I can't believe that anyone would make the following statement
without joking or being incredibly naive:

> " As you might have surmised from my ontology paper, I have
> a single uniform coding system for all concepts. In any given knowledge
> base, absolutely everything (Content and Tool components) is characterized
> by a (Concept Sub-Code) and (Instance Sub-Code) pair."

That statement is at the level of claiming that C is a uniform coding
language for all knowledge because "absolutely everything (content
and tool components) can be characterized by a pair of pointers
to a (concept sub-code) and (instance sub-code) pair."

In 1957, Silvio Ceccato used the IBM 650 (a computer with a rotating
drum for memory) to represent all relations by a pair of pointers
and a code for the type of relation.  Conveniently, the 650 had
a word length of 10 decimal digits, of which the first two were the
code, and the next 8 were two pairs of 4-digit pointers.  But we have
come a long way from that "uniform representation".

If anyone is interested in a survey of the kinds of things that have
been done with semantic networks over the past 40+ years, I would
recommend my article:

    http://www.jfsowa.com/pubs/semnetw.htm

This article is a draft that will eventually be published in the
forthcoming _Encyclopedia of Cognitive Science_:

    http://www.macmillanonline.net/Science/ecs.htm

Following is the opening section.

John Sowa
__________________________________________________________________________

                            Semantic Networks

                               John F. Sowa

A semantic network or net is a graphic notation for representing
knowledge in patterns of interconnected nodes and arcs.  Computer
implementations of semantic networks were first developed for artificial
intelligence and machine translation, but earlier versions have long
been used in philosophy, psychology, and linguistics.

What is common to all semantic networks is a declarative graphic
representation that can be used either to represent knowledge or to
support automated systems for reasoning about knowledge.  Some versions
are highly informal, but other versions are formally defined systems of
logic.  Following are six of the most common kinds of semantic networks,
each of which is discussed in detail in one section of this article.

  1. Definitional networks emphasize the subtype or is-a relation between
     a concept type and a newly defined subtype.  The resulting network,
     also called a generalization or subsumption hierarchy, supports the
     rule of inheritance for copying properties defined for a supertype
     to all of its subtypes.  Since definitions are true by definition,
     the information in these networks is often assumed to be necessarily
     true.

  2. Assertional networks are designed to assert propositions.  Unlike
     definitional networks, the information in an assertional network is
     assumed to be contingently true, unless it is explicitly marked with
     a modal operator.  Some assertional netwoks have been proposed as
     models of the conceptual structures underlying natural language
     semantics.

  3. Implicational networks use implication as the primary relation for
     connecting nodes.  They may be used to represent patterns of
beliefs,
     causality, or inferences.

  4. Executable networks include some mechanism, such as marker passing
     or attached procedures, which can perform inferences, pass messages,
     or search for patterns and associations.

  5. Learning networks build or extend their representations by acquiring
     knowledge from examples.  The new knowledge may change the old
     network by adding and deleting nodes and arcs or by modifying
     numerical values, called weights, associated with the nodes and
     arcs.

  6. Hybrid networks combine two or more of the previous techniques,
     either in a single network or in separate, but closely interacting
     networks.

Some of the networks have been explicitly designed to implement
hypotheses about human cognitive mechanisms, while others have been
designed primarily for computer efficiency.  Sometimes, computational
reasons may lead to the same conclusions as psychological evidence.  The
distinction between definitional and assertional networks, for example,
has a close parallel to Tulving's (1972) distinction between semantic
memory and episodic memory.

Network notations and linear notations are both capable of expressing
equivalent information, but certain representational mechanisms are
better suited to one form or the other.  Since the boundary lines are
vague, it is impossible to give necessary and sufficient conditions that
include all semantic networks while excluding other systems that are not
usually called semantic networks.  Section 7 of this article discusses
the syntactic mechanisms used to express information in network
notations
and compares them to the corresponding mechanisms used in linear
notations.

Kal Ahmed wrote:
>
> > Kal Ahmed wrote:
> > [...]
> > > Perhaps I'm way off here, but don't people use the ISO codes
> > precisely to
> > > unambiguously represent the country/language ?
> >
> > Yes they do. But the codes themselves are described as codes, not as
> > languages or countries. ISO and IETF disabuse themselves of that semantic
> > problem and basically say that if you want to refer to "German" using
> > our code, you're free to do so, but not with the idea that you're
> > doing any more than that. It's sort of like when you buy something and
> > they give you that multipage legal disclaimer that you toss away without
> > reading. "Please don't hold us responsible for the problems that might
> > occur if two or more people use our codes to represent things like
> > languages and countries that we know don't have any firm semantic
> > boundaries, that if you use 'English' between yourselves you agree
> > that it's just a proxy for whatever you think 'English' is."
> >
> > My updated topic maps do not use "ISO Language Code" as the typing
> > topic, but of the updates simply describes the PSIs as instances of
> > languages or countries, since that semantic (as I've been for quite
> > awhile now) been trying to point out would be a fallacy.

Sorry, I mistyped. I meant to say that the updated topic maps do not
use "ISO Language Codes" as a typing topic, as simply typing them as
languages or countries would be problematic. Just as with ISO or IETF,
they are "topics to represent countries or languages." The "to represent"
is extremely important.

> Hmm - thats a shame, because I think that a set of PSIs for both the
> countries themselves would be a major step forward in permitting consistent
> topic map merging. Of course, I fully agree with your assertion the codes
> are not the countries, but PSIs *based on* those codes could serve as a
> starting point for such a topic map (and yes, I do realise that there are
> geo-political issues and historical boundary changes which have to be
> resolved or side-stepped by such a topic map).

That's precisely what the PSIs do. They take the same form as an ISO
code. Those people willing to use them take the same risks as using
the codes. They are proxies for whatever an author believes "Germany"
or "English" means to them.

> > As for debating this much further, I've been trying to make plain that
> > I want this process to occur in the group. I find Lars Marius' tactic
> > of posting this now perhaps suits his agenda but is simply requiring
> > that I either take part in an unformed discussion now or bow out
>
> I'm sorry if you feel that I am uninformed. I believe that I am only
> uninformed about your perspective on language.xtm and coutry.xtm, hence my
> question.

"Unformed" not "uninformed." I hardly think you uninformed.

> > and wait until we have a proper process under which to discuss this
> > and come to real resolution over the issues. I prefer to bow out and
> > wait the several weeks until Montreal.
>
> That is a pity as I am not able to attend the meeting at Montreal and I feel
> that I would like to contribute to this discussion. If you are not willing
> to do so then there is no point in banging my head against this particular
> wall. I have plenty of others that also require precious forehead skin.

Don't say that last sentence after too many beers. Someone might get
the wrong idea. :-)

Murray

...........................................................................
Murray Altheim, SGML/XML Grease Monkey  <mailto:murray.altheim&#64;sun.com>
XML Technology Center
Sun Microsystems, 1601 Willow Rd., MS UMPK17-102, Menlo Park, CA 94025

      i am going to see if i cannot reform insects in general
      i have constituted myself a missionary extraordinary
      and minister plenipotentiary and entomological to bring
      idealism to the little struggling brothers -- archy (1927)

-----Original Message-----
From: John F. Sowa [mailto:sowa@...]
Sent: Thursday, July 26, 2001 6:13 AM
To: Paul Stephen Prueitt
Cc: OntologyStream; Jon Awbrey; W.M. Jaworski; Xtm-Wg; Steve Pepper; Steven R. Newcomb; Brian (Bo) Newman; Stand! Unfold! Ontology!; Arisbe; Robert Shaw; Ray Fergerson; 'Com-Prac
Subject: [xtm-wg] Re: Inquiry Into Inquiry

Paul Stephen Prueitt wrote:

> I ask John Sowa to make a comment about the fact that Tulving has declared
> his former views, regarding the distinction between semantic and episodic
> memory, as being a distinction that has been found to be lacking.  Are we
> simply to ignore this history?

I never regarded it as a crucial distinction upon which everything
else stands or falls.  In AI, the distinction between the definitional
networks and the assertional networks has been a useful way of dividing
up the task and organizing the various pieces of the puzzle.

And in fact, the approach that I have been developing in recent years
can be interpreted in different ways, some of which could be viewed
as supporting either position.

> If science cannot walk away from an establish paradigm, when evidence sets
> it aside, then why have a notion of falsification at all?

There were never any claims that could be or have been falsified.
The AI systems use both definitions and assertions.  Any particular
proposition that follows from the conjunction of both would still
be derived whether the two kinds of information were stored separately
or together.

There are several distinct fields:

1. Neuropsychology:  How do human and animal brains work?

2. AI science:  What are the logical foundations of learning
    and reasoning?

3. AI engineering:  How does one build intelligent machines that
    learn and reason effectively?

Each of these fields has had some influence on each of the others,
but no particular result from any one of them necessarily contradicts
any result of any of the others.  For example, you may have logical
theories that are inefficient in any neural or silicon implementation.
Or you could find others that are very good candidates for one, but not
the other.

John Sowa

To Post a message, send it to:   xtm-wg@yahooGroups.com

To Unsubscribe, send a blank message to: xtm-wg-unsubscribe@yahooGroups.com

Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.

> Kal Ahmed wrote:
> [...]
> > Perhaps I'm way off here, but don't people use the ISO codes
> precisely to
> > unambiguously represent the country/language ?
>
> Yes they do. But the codes themselves are described as codes, not as
> languages or countries. ISO and IETF disabuse themselves of that semantic
> problem and basically say that if you want to refer to "German" using
> our code, you're free to do so, but not with the idea that you're
> doing any more than that. It's sort of like when you buy something and
> they give you that multipage legal disclaimer that you toss away without
> reading. "Please don't hold us responsible for the problems that might
> occur if two or more people use our codes to represent things like
> languages and countries that we know don't have any firm semantic
> boundaries, that if you use 'English' between yourselves you agree
> that it's just a proxy for whatever you think 'English' is."
>
> My updated topic maps do not use "ISO Language Code" as the typing
> topic, but of the updates simply describes the PSIs as instances of
> languages or countries, since that semantic (as I've been for quite
> awhile now) been trying to point out would be a fallacy.
>

Hmm - thats a shame, because I think that a set of PSIs for both the
countries themselves would be a major step forward in permitting consistent
topic map merging. Of course, I fully agree with your assertion the codes
are not the countries, but PSIs *based on* those codes could serve as a
starting point for such a topic map (and yes, I do realise that there are
geo-political issues and historical boundary changes which have to be
resolved or side-stepped by such a topic map).

> As for debating this much further, I've been trying to make plain that
> I want this process to occur in the group. I find Lars Marius' tactic
> of posting this now perhaps suits his agenda but is simply requiring
> that I either take part in an unformed discussion now or bow out

I'm sorry if you feel that I am uninformed. I believe that I am only
uninformed about your perspective on language.xtm and coutry.xtm, hence my
question.

> and wait until we have a proper process under which to discuss this
> and come to real resolution over the issues. I prefer to bow out and
> wait the several weeks until Montreal.
>

That is a pity as I am not able to attend the meeting at Montreal and I feel
that I would like to contribute to this discussion. If you are not willing
to do so then there is no point in banging my head against this particular
wall. I have plenty of others that also require precious forehead skin.

Cheers,

Kal

¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤

Paul Stephen Prueitt wrote (PSP):
W.M. Jaworski wrote (WMJ):
Jon Awbrey wrote (JA):

PSP: I am not sure which of the forums this will be posted to.
      I have left the Topic Maps forum because there is no response
      to the issues that I have very carefully brought up.

PSP: Arisbe and "Stand! Unfold! Ontology!" are unknown to me.
      I post this to com-prac because the discussion is essential
      to defending community of practice work from those who insist
      that the IT is the most important part of collaborative tools.

Information about the Arisbe List, whose charge is something on the order
of "e-Peirce!" -- to ask how the pragmatudes of Peirce and the webwork of
our brave modworld may best serve each other as we lurk into the future --
can be found at:

http://stderr.org/cgi-bin/mailman/listinfo/arisbe
http://stderr.org/pipermail/arisbe

The trinicker "Stand! Unfold! Ontology!" is just my address book niche-name
for the "ONTology" Sublist of the IEEE "Standard Upper Ontology" (SUO) Agora
and Working Group.  There are actually four subfora associated with this body,
ONT being the Brand X Generic one.  Information about these e-fora is here:

http://suo.ieee.org
http://suo.ieee.org/refs.html
http://suo.ieee.org/links.html

http://suo.ieee.org/email
http://suo.ieee.org/ontology
http://suo.ieee.org/suo-ce
http://suo.ieee.org/suo-kif

PSP: There is the computer science and the human science, and these two
viewpoints
      must be balanced if **Value Propositions** are to pay out a return on
investment.

***

JA: Okay, maybe, but I wasted far too many years playing the viscous circuits of
     declarative/procedural volleyball to want to go through that again, which is
     why I left cogsci for a better refereed net.  Provisionally, though, if you
     declare "declarative" to mean something like the kind of fully critical and
     more or less reflective knowledge that it takes, in the ideal, to write out
     an explicit axiomset for a domain, a grammar for a language, or a program
     for a recursive function, then I think that I can go along with that.

PSP: I also feel a waste of my time and of the time of our society
      on this declarative/procedural distinction.

PSP: I read through John Sowa scholarly work that he posted at:

JFS: http://www.jfsowa.com/pubs/semnetw.htm

PSP: and the history is familiar.  But as I read this history,
      I keep asking the question, "what is missing here?"

PSP: Tulving is referenced on page 1, but this is the 1972 work that Tulving
      specifically deconstructs, and disavows, in the 1980s and the 1990s.

The 80's were like that, yeah they were.

PSP: In the volume edited by Schacter and Tulving ('Memory' 1993)
      the best minds in human memory research essentially say that
      the earlier distinction between semantic memory and episodic
      memory can not be supported by the modern empirical work.

PSP: quote from Schacter and Tulving page 2:

S&T: | The fact that concern with the neural mechanisms underlying psychological
manifestations
      | of memory and knowledge was absent in the earliest thought about
different forms of memory
      | is not surprising.  On the one hand, there was little understanding of
the brain at the time,
      | and thus the neural influence was necessarily absent.  On the other hand,
the doctrine of
      | associationism held almost universal sway over philosophical and
psychological thinking
      | about memory, rendering any kind of physiologizing superfluous. 
Moreover, the associative
      | doctrine was dominated by the idea that all expressions of memory could
be attributed to the
      | functioning of a single associative mechanism, an idea that is still
around even today.

PSP: What is missing is the modern view of cognitive neuroscience.
      I do not mean Pat Churchland's work or other work that is
      developed (strongly supported by NSF) in a specific means
      to support the hard AI position, but rather the core
      experimental work on understanding the biology of
      cognition and behavior.

And here I had not thought you were such a modernist
that you would let youself to fall into that modern,
all too modern habit of using the holy word "modern"
in lieu of any argument as to why a teaching is true.

PSP: This is a difficult discussion with the strong AI camp.
      The discussion brings to bear the modern research on
      human memory and anticipation as a means to try to
      deconstruct the strong AI position.

It is generaly considered a clash of genres and quite unfashionable
to resort to both the word "modern" and the word "deconstruct" in
one and the same text.

PSP: As Tulving and Schacter state, there is deeper evidence about memory
      and this deeper evidence does not support specific previous viewpoints --
      from which both strong AI and much of connectionism is derived.

PSP: It is a consuming debate, requiring continual review of books
      and the writing and rewriting of e-forum communications.

PSP: This process might go on for a few more days.
      http://groups.yahoo.com/group/OntologyStream
      is an open forum, where those wishing to
      discuss this may gather.

PSP: I, and my colleagues, want to try to establish a new position (change a
mind)
      but realize that we may just increase our own understanding of the
entrenchment
      of strong AI.  In phone discussions on this, it is clear that entrenched
positions
      will not allow principled argument -- and often fail back into claims that
some harm
      is being done to the good paradigm.  One is not allowed to bring up the
fact that there
      are no intelligent computer programs out there.  There are no computer
programs out there
      that are a "they".  There are no computer programs out there that "know".

I know, I know.  But what makes you think you know that I am not a machine?
I think I know.  But what makes you think you know that I am not a program?
I know I think.  But what makes you think you know that I am not a poem?

PSP: The language used is simply wrong and misleading.  By pointing this out,
      my camp does not begin talking about "God" or a theory of everything,
      we are just pointing out a clear fact that the AI community continues
      to be in denial about.  The behavior of those entrenched minds is
      scandalous to science as a discipline.

PSP: One of the reasons why I have spent most of my life looking to develop a
bead game
      that converts topics in a e-forum, or other discussion, into a machine
knowledge
      ontology is so that the slight of hand used by the entrenched camps can be
      revealed for what they are.  The **Value proposition** is that by shining
      light on why the current investment in IT technology are not paying off,
      society might benefit form a new epoch.

{[(<| Or, better yet, a new epoche. |>)]}

PSP: Important problems, like Computer Intrusion Detection (CID) or Situational
Decision Support,
      just require an easier interface between human communities and the data
structures.  We need
      less strong AI religion and more agility in the data structures.  And this
interface requires
      an understanding of the fundamental difference between a finite state
machine and a human mind
      or a human community.  In the CID world the situation is special, in that
the attackers mush act
      via the computer ... and so AI done right has a chance of doing much more
than it is doing now.
      But ... there are still the problem that AI is not non-stationary in the
same way as a natural
      system.

PSP: http://www.bcngroup.org/area3/pprueitt/kmbook/Chapter2.htm

PSP: I ask John Sowa to make a comment about the fact that Tulving has declared
his former views,
      regarding the distinction between semantic and episodic memory, as being a
distinction that
      has been found to be lacking.  Are we simply to ignore this history?

History shows that ignoring history
becomes easier and easier over time.

PSP: If science cannot walk away from an establish paradigm,
      when evidence sets it aside, then why have a notion of
      falsification at all?

Good question.

Post*Sincerely Yours,

Post*Modern Jon

P.S.  Post*Modern = Modern + PostModern + PostPostModern + PostPostPostModern +
...
J.A.

¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤~~~~~~~~~¤

Kal Ahmed wrote:
[...]
> Perhaps I'm way off here, but don't people use the ISO codes precisely to
> unambiguously represent the country/language ?

Yes they do. But the codes themselves are described as codes, not as
languages or countries. ISO and IETF disabuse themselves of that semantic
problem and basically say that if you want to refer to "German" using
our code, you're free to do so, but not with the idea that you're
doing any more than that. It's sort of like when you buy something and
they give you that multipage legal disclaimer that you toss away without
reading. "Please don't hold us responsible for the problems that might
occur if two or more people use our codes to represent things like
languages and countries that we know don't have any firm semantic
boundaries, that if you use 'English' between yourselves you agree
that it's just a proxy for whatever you think 'English' is."

My updated topic maps do not use "ISO Language Code" as the typing
topic, but of the updates simply describes the PSIs as instances of
languages or countries, since that semantic (as I've been for quite
awhile now) been trying to point out would be a fallacy.

As for debating this much further, I've been trying to make plain that
I want this process to occur in the group. I find Lars Marius' tactic
of posting this now perhaps suits his agenda but is simply requiring
that I either take part in an unformed discussion now or bow out
and wait until we have a proper process under which to discuss this
and come to real resolution over the issues. I prefer to bow out and
wait the several weeks until Montreal.

Murray

...........................................................................
Murray Altheim                         <mailto:murray.altheim&#x40;sun.com>
XML Technology Center
Sun Microsystems, Inc., MS MPK17-102, 1601 Willow Rd., Menlo Park, CA 94025

       In the evening
       The rice leaves in the garden
       Rustle in the autumn wind
       That blows through my reed hut.  -- Minamoto no Tsunenobu

Murray wrote:

> Lars Marius Garshol wrote:
> >
> > * H. Holger Rath
> > |
> > | Thanks for this valuable work. I hope we could approve both
> > | in Montréal.
> >
> > So do I. :-)
> >
> > | We discussed this in length in Austin and the group drew to the
> > | conclusion that we cannot define what the language and what the
> > | country subjects are. Furthermore, the subjects defined in
> > | language.xtm and country.xtm are no languages and no contries; the
> > | subjects are the codes ISO defined for some languages and countries.
> >
> > In that case there is no point in having these PSI sets.
> >
> > I can understand that people feel uneasy about "standardizing the
> > languages of the world", but this has already been done by ISO 639,
> > and if anyone has a problem with that they should take it up with the
> > responsible ISO committee and DIN as the maintenance agency.
>
> No, this was the crucial point. They did *not* standardize the languages
> and countries of the world, they standardized a list of their *names*
> as well as a set of codes as proxies for those names. Look at the name
> of each standard. Both ISO and IETF always call them "codes for the
> representation of languages", "Tags for the identification of languages,"
> etc. It's fallacious to do otherwise, and the current language and
> country topic maps do exactly the same thing, by conscious design. I
> remember arguing with Steve Pepper about this same issue. He's wrong
> and so is Lars Marius, unless we want to depart from ISO's usage.
> People can use the PSIs exactly as they would the ISO codes, and this
> is appropriate.
>

Perhaps I'm way off here, but don't people use the ISO codes precisely to
unambiguously represent the country/language ?

> > In i18n.ltm I have an association that says that German is spoken in
> > Germany. I can't make German and Germany instances of codes, because
> > the one code is not spoken in the other; in other words, I would be
> > saying something different from what I mean. The same issue of course
> > appears in a huge number of topic maps.
>
> Agreed.
>
> > So if the topics in language.xtm are not instances of languages they
> > are useless. The reason we want to have them is that we want to be
> > able to talk about languages. If we end up having topic maps that make
> > statements about language codes we have achieved nothing.
>
> No, the PSIs established by these topic maps serve as proxies for
> the subjects, just as does any PSI. When people use the PSI they
> are agreeing in the same way as when they use the ISO code.
>

But aren't they agreeing on the country/language, not on the code for it ?

> > Furthermore, the current version of language.xtm says that the
> > language code 'de' has two names:
> >
> >   'de'
> >   'German'
> >
> > I think it is reasonable to claim that the name of the language code
> > is 'de', but to claim that the code has the name 'German' is absurd.
> > There is a language, known by the names 'German', 'Deutsch', 'tysk',
> > and many others, as well as by the code 'de', but the code itself has
> > no name but itself.
>
> This is either a red herring or you've not read the topic map very well.
> The 'de' string is scoped by being an alpha-2 code. The 'German' string
> is scoped by "English" meaning it's an identifier for anything German
> when you're speaking English. Let's be clear here.
>

This sounds to me like you are saying that the subject of the topic is
"German" and not "The ISO two-letter code for German", which would a more
precise form of English if the subject of the topic was the code and not the
language. So this confuses my in conjunction with your previous statements.

> > In short, I can't see how these topic maps can be of any use to anyone
> > unless the topics in them are instances of 'country' and 'language',
> > respectively. Surely that must be the point of the whole exercise?
>
> Your argument holds no water; ISO and IETF have been doing precisely
> the same thing as these topic maps for many years. Unless we believe
> that TopicMaps.Org can be more authoritive than ISO we should steer
> clear of any such pretense of authoritarianism.
>

Perhaps I am being dim, but I feel that I need to understand this clearly
before expressing any more opinions on either language.xtm file, so please
try and help me on this:

Are the (majority of) topics in language.xtm (for now, lets just concentrate
on that one), as published on the TopicMaps.org website, intended to
represent the concept of "Languages" or the concept of "ISO Codes for
Languages" ?

Cheers,

Kal

Lars Marius Garshol wrote:
>
> * H. Holger Rath
> |
> | Thanks for this valuable work. I hope we could approve both
> | in Montréal.
>
> So do I. :-)
>
> | We discussed this in length in Austin and the group drew to the
> | conclusion that we cannot define what the language and what the
> | country subjects are. Furthermore, the subjects defined in
> | language.xtm and country.xtm are no languages and no contries; the
> | subjects are the codes ISO defined for some languages and countries.
>
> In that case there is no point in having these PSI sets.
>
> I can understand that people feel uneasy about "standardizing the
> languages of the world", but this has already been done by ISO 639,
> and if anyone has a problem with that they should take it up with the
> responsible ISO committee and DIN as the maintenance agency.

No, this was the crucial point. They did *not* standardize the languages
and countries of the world, they standardized a list of their *names*
as well as a set of codes as proxies for those names. Look at the name
of each standard. Both ISO and IETF always call them "codes for the
representation of languages", "Tags for the identification of languages,"
etc. It's fallacious to do otherwise, and the current language and
country topic maps do exactly the same thing, by conscious design. I
remember arguing with Steve Pepper about this same issue. He's wrong
and so is Lars Marius, unless we want to depart from ISO's usage.
People can use the PSIs exactly as they would the ISO codes, and this
is appropriate.

> In i18n.ltm I have an association that says that German is spoken in
> Germany. I can't make German and Germany instances of codes, because
> the one code is not spoken in the other; in other words, I would be
> saying something different from what I mean. The same issue of course
> appears in a huge number of topic maps.

Agreed.

> So if the topics in language.xtm are not instances of languages they
> are useless. The reason we want to have them is that we want to be
> able to talk about languages. If we end up having topic maps that make
> statements about language codes we have achieved nothing.

No, the PSIs established by these topic maps serve as proxies for
the subjects, just as does any PSI. When people use the PSI they
are agreeing in the same way as when they use the ISO code.

> Furthermore, the current version of language.xtm says that the
> language code 'de' has two names:
>
>   'de'
>   'German'
>
> I think it is reasonable to claim that the name of the language code
> is 'de', but to claim that the code has the name 'German' is absurd.
> There is a language, known by the names 'German', 'Deutsch', 'tysk',
> and many others, as well as by the code 'de', but the code itself has
> no name but itself.

This is either a red herring or you've not read the topic map very well.
The 'de' string is scoped by being an alpha-2 code. The 'German' string
is scoped by "English" meaning it's an identifier for anything German
when you're speaking English. Let's be clear here.

> In short, I can't see how these topic maps can be of any use to anyone
> unless the topics in them are instances of 'country' and 'language',
> respectively. Surely that must be the point of the whole exercise?

Your argument holds no water; ISO and IETF have been doing precisely
the same thing as these topic maps for many years. Unless we believe
that TopicMaps.Org can be more authoritive than ISO we should steer
clear of any such pretense of authoritarianism.

Murray

...........................................................................
Murray Altheim                         <mailto:murray.altheim&#x40;sun.com>
XML Technology Center
Sun Microsystems, Inc., MS MPK17-102, 1601 Willow Rd., Menlo Park, CA 94025

       In the evening
       The rice leaves in the garden
       Rustle in the autumn wind
       That blows through my reed hut.  -- Minamoto no Tsunenobu

Kal Ahmed wrote:
>
> Holger wrote:
>
> > Hi Lars Marius,
> >
> > Thanks for this valuable work. I hope we could approve both
> > in Montréal.

I know I've already done my complaining about the idea of having two
sets of language and country topic maps to deal with, and consider it
really poor form to create a competition where none need exist. But
since Lars Marius is as he says "willing to risk" the downside of that,
the group must now decide between the ones I've authored and these
new ones.

Since my revised language and country topic maps do everything that
Lars Marius' does (as far as I can see), also maps in the UN and
MARC codes, and doesn't do things I disagree with (such as remove
the necessary 'xml:base' attribute or make the XTM DTD URI absolute,
which will just cause parsing headaches), I will of course vote against
approval. This all seems pretty stupid as a "process" though. Do I
now have to expect that someone will duplicate everything else I'm
doing? I don't consider this "competition in the marketplace," nor
do I appreciate what should be a cooperative venture in a standards
activity being turned into a competition. Lars Marius has justified
this as his requirement on getting the work done before Montreal, but
I don't buy the justification. I've already stated that he can use
the PSIs established by the current topic maps, as they are opaque
containers for those PSIs.

Put it this way: I've put *many* hours into preparing the updates to
the languages and countries topic maps. They will be ready and delivered
to the group when it begins to discuss this issue. I will *certainly*
vote against using someone else's versions on principle. And should
the group vote to use someone else's, I would be pretty pissed off.
Is this how we should be working cooperatively together under our new
organization? Is this how our process and activity should be run? I
hope not, as I'd certainly reconsider how much time and commitment I'd
put into such an activity in the future. We either work together or we
fail. And as I've said, I don't consider this announcement very productive
towards regaining any trust amongst a group whose trust is at a low
point. I am strongly against spending the time on this prior to having
some process under which we can mediate disagreement. And I disagree
strongly with many of the decisions made in these new topic maps. Very
strongly, both from lexical, syntactic and semantic standpoints.

But as he says, he's willing to risk it.

[...]
> I wasn't present at the Austin meeting, but in looking at these proposed
> topic maps, I would be of the opinion that the topics are surrogates for
> countries and languages. It is the names that *are* the two and three letter
> code (as well as the English and French names), but the topics are suitable
> surrogates for countries and languages, and those codes are *used as*
> subject indicators, but IMHO the subject being indicated is the
> country/language in question.

This is precisely how the ISO codes are described and used, and the
current language and country topic maps were authored with the same
idea in mind. There is no need to break with this tradition, because
use of a surrogate for a language or country is all you can do anyway.



Murray

...........................................................................
Murray Altheim                         <mailto:murray.altheim&#x40;sun.com>
XML Technology Center
Sun Microsystems, Inc., MS MPK17-102, 1601 Willow Rd., Menlo Park, CA 94025

       In the evening
       The rice leaves in the garden
       Rustle in the autumn wind
       That blows through my reed hut.  -- Minamoto no Tsunenobu

* H. Holger Rath
|
| Thanks for this valuable work. I hope we could approve both
| in Montréal.

So do I. :-)

| We discussed this in length in Austin and the group drew to the
| conclusion that we cannot define what the language and what the
| country subjects are. Furthermore, the subjects defined in
| language.xtm and country.xtm are no languages and no contries; the
| subjects are the codes ISO defined for some languages and countries.

In that case there is no point in having these PSI sets.

I can understand that people feel uneasy about "standardizing the
languages of the world", but this has already been done by ISO 639,
and if anyone has a problem with that they should take it up with the
responsible ISO committee and DIN as the maintenance agency.

In i18n.ltm I have an association that says that German is spoken in
Germany. I can't make German and Germany instances of codes, because
the one code is not spoken in the other; in other words, I would be
saying something different from what I mean. The same issue of course
appears in a huge number of topic maps.

So if the topics in language.xtm are not instances of languages they
are useless. The reason we want to have them is that we want to be
able to talk about languages. If we end up having topic maps that make
statements about language codes we have achieved nothing.

Furthermore, the current version of language.xtm says that the
language code 'de' has two names:

   'de'
   'German'

I think it is reasonable to claim that the name of the language code
is 'de', but to claim that the code has the name 'German' is absurd.
There is a language, known by the names 'German', 'Deutsch', 'tysk',
and many others, as well as by the code 'de', but the code itself has
no name but itself.

In short, I can't see how these topic maps can be of any use to anyone
unless the topics in them are instances of 'country' and 'language',
respectively. Surely that must be the point of the whole exercise?

--Lars M.

Paul Stephen Prueitt wrote:

> I ask John Sowa to make a comment about the fact that Tulving has declared
> his former views, regarding the distinction between semantic and episodic
> memory, as being a distinction that has been found to be lacking.  Are we
> simply to ignore this history?

I never regarded it as a crucial distinction upon which everything
else stands or falls.  In AI, the distinction between the definitional
networks and the assertional networks has been a useful way of dividing
up the task and organizing the various pieces of the puzzle.

And in fact, the approach that I have been developing in recent years
can be interpreted in different ways, some of which could be viewed
as supporting either position.

> If science cannot walk away from an establish paradigm, when evidence sets
> it aside, then why have a notion of falsification at all?

There were never any claims that could be or have been falsified.
The AI systems use both definitions and assertions.  Any particular
proposition that follows from the conjunction of both would still
be derived whether the two kinds of information were stored separately
or together.

There are several distinct fields:

  1. Neuropsychology:  How do human and animal brains work?

  2. AI science:  What are the logical foundations of learning
     and reasoning?

  3. AI engineering:  How does one build intelligent machines that
     learn and reason effectively?

Each of these fields has had some influence on each of the others,
but no particular result from any one of them necessarily contradicts
any result of any of the others.  For example, you may have logical
theories that are inefficient in any neural or silicon implementation.
Or you could find others that are very good candidates for one, but not
the other.

John Sowa

Holger wrote:

> Hi Lars Marius,
>
> Thanks for this valuable work. I hope we could approve both
> in Montréal.
>
> But I have a comment (in fact dealing with a simlar issue).
>
> You wrote:
> > ... snip ...
> >
> > - the topic for language-code has been replaced by a topic for
> >    language, so that there now exists a well-defined PSI for language
> >
> > ... snip ...
> >
> >  - the topic for country-code has been replaced by a topic for
> >    country, so that there now exists a well-defined PSI for country
>
> We discussed this in length in Austin and the group drew to the conclusion
> that we cannot define what the language and what the country subjects are.
> Furthermore, the subjects defined in language.xtm and country.xtm
> are no languages
> and no contries; the subjects are the codes ISO defined for some languages
> and countries. That is why we had the topics language-code and
> country-code
> and my feeling is that they are correct (from a philosophical
> point of view).
>

I wasn't present at the Austin meeting, but in looking at these proposed
topic maps, I would be of the opinion that the topics are surrogates for
countries and languages. It is the names that *are* the two and three letter
code (as well as the English and French names), but the topics are suitable
surrogates for countries and languages, and those codes are *used as*
subject indicators, but IMHO the subject being indicated is the
country/language in question.

Cheers,

Kal

> Sorry, to tell you that at this late point in time, but I thought
> SteveP would remind you.
>
> Cheers,
> --Holger
>
> --
> Dr. H. Holger Rath
> - Director Research & Development -
> empolis * GmbH
> Bertelsmann MOHN Media Group
>
>
> To Post a message, send it to:   xtm-wg@yahooGroups.com
>
> To Unsubscribe, send a blank message to:
> xtm-wg-unsubscribe@yahooGroups.com
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>
>
>

I am not sure which of the forums this will be posted to.  I have left the
Topic Maps forum because there is no response to the issues that I have very
carefully brought up.

Arisbe and Stand! Unfold! and Ontology! are unknown to me.  I post this to
com-prac because the discussion is essential to defending community of
practice work from those who insist that the IT is the most important part
of collaborative tools.

There is the computer science and the human science, and these two
viewpoints must be balanced if **Value Propositions** are to pay out a
return on investment.

***

Jaworski said:

"Okay, maybe, but I wasted far too many years playing the viscous circuits
of
declarative/procedural volleyball to want to go through that again, which is
why I left cogsci for a better refereed net."

****

I also feel a waste of my time and of the time of our society on this
declarative/procedural distinction.


I read through John Sowa scholarly work that he posted at;

    http://www.jfsowa.com/pubs/semnetw.htm


and the history is familiar.  But as I read this history, I keep asking the
question, "what is missing here?"

Tulving is referenced on page 1, but this is the 1972 work that Tulving
specifically deconstructs, and disavows, in the 1980s and the 1990s.  In the
volume edited by Schacter and Tulving (Memory 1993) the best minds in human
memory research essentially say that the earlier distinction between
semantic memory and episodic memory can not be supported by the modern
empirical work.

"quote from Schacter and Tulving page 2 -"

" The fact that concern with the neural mechanisms underlying psychological
manifestations of memory and knowledge was absent in the earliest thought
about different forms of memory is not surprising.  On the one hand, there
was little understanding of the brain at the time, and thus the neural
influence was necessarily absent. On the other hand, the doctrine of
associationism held almost universal sway over philosophical and
psychological thinking about memory, rendering any kind of physiologizing
superfluous.  Moreover, the associative doctrine was dominated by the idea
that all expressions of memory could be attributed to the functioning of a
single associative mechanism, an idea that is still around even today. "

***

What is missing is the modern view of cognitive neuroscience.  I do not mean
Pat Churchland's work or other work that is developed (strongly supported by
NSF) in a specific means to support the hard AI position, but rather the
core experimental work on understanding the biology of cognition and
behavior.

This is a difficult discussion with the strong AI camp.  The discussion
brings to bear the modern research on human memory and anticipation as a
means to try to deconstruct the strong AI position.  As Tulving and Schacter
state, there is deeper evidence about memory and this deeper evidence does
not support specific previous viewpoints - from which both strong AI and
much of connectionism is derived.

It is a consuming debate, requiring continual review of books and the
writing and rewriting of e-forum communications.

This process might go on for a few more days.

http://groups.yahoo.com/group/OntologyStream is an open forum.. where those
wishing to discussion this may gather.

I, and my colleagues, want to try to establish a new position (change a
mind) but realize that we may just increase our own understanding of the
entrenchment of strong AI.  In phone discussions on this, it is clear that
entrenched positions will not allow principled argument - and often fail
back into claims that some harm is being done to the good paradigm.  One is
not allowed to bring up the fact that there are no intelligent computer
programs out there.  There are no computer programs out there that are a
"they".  There are no computer programs out there that "know".

The language used is simply wrong and misleading.  By pointing this out, my
camp does not begin talking about "God" or a theory of everything, we are
just pointing out a clear fact that the AI community continues to be in
denial about.  The behavior of those entrenched minds is scandalous to
science as a discipline.

One of the reasons why I have spent most of my life looking to develop a
bead game that converts topics in a e-forum, or other discussion, into a
machine knowledge ontology is so that the slight of hand used by the
entrenched camps can be revealed for what they are.  The **Value
proposition** is that by shining light on why the current investment in IT
technology are not paying off, society might benefit form a new epoch.

Important problems, like Computer Intrusion Detection (CID) or Situational
Decision Support, just require an easier interface between human communities
and the data structures.  We need less strong AI religion and more agility
in the data structures.  And this interface requires an understanding of the
fundamental difference between a finite state machine and a human mind or a
human community.  In the CID world the situation is special, in that the
attackers mush act via the computer.. and so AI done right has a chance of
doing much more than it is doing now.  But ...  there are still the problem
that AI is not non-stationary in the same way as a natural system.

http://www.bcngroup.org/area3/pprueitt/kmbook/Chapter2.htm

I ask John Sowa to make a comment about the fact that Tulving has declared
his former views, regarding the distinction between semantic and episodic
memory, as being a distinction that has been found to be lacking.  Are we
simply to ignore this history?

If science cannot walk away from an establish paradigm, when evidence sets
it aside, then why have a notion of falsification at all?

Hi Lars Marius,

Thanks for this valuable work. I hope we could approve both
in Montréal.

But I have a comment (in fact dealing with a simlar issue).

You wrote:
> ... snip ...
>
> - the topic for language-code has been replaced by a topic for
>    language, so that there now exists a well-defined PSI for language
>
> ... snip ...
>
>  - the topic for country-code has been replaced by a topic for
>    country, so that there now exists a well-defined PSI for country

We discussed this in length in Austin and the group drew to the conclusion
that we cannot define what the language and what the country subjects are.
Furthermore, the subjects defined in language.xtm and country.xtm are no
languages
and no contries; the subjects are the codes ISO defined for some languages
and countries. That is why we had the topics language-code and country-code
and my feeling is that they are correct (from a philosophical point of view).

Sorry, to tell you that at this late point in time, but I thought
SteveP would remind you.

Cheers,
--Holger

--
Dr. H. Holger Rath
- Director Research & Development -
empolis * GmbH
Bertelsmann MOHN Media Group

thanks for the recent discussion of TMs with a slant to the
indexing/librarian community, by glenda brown & murray altheim.

as an information scientist in knowledge organization with a certain
sympathy for TMs, i think i understand your concerns.
let me add some sketchy remarks (they became longer than i intended):

glenda had asked three important questions:

1. Do you know of good introductory material?
---------------------------------------------
    although we are now far better off than at xml europe 2000 in paris,
    this is still a desideratum. what i can offer is material about
    how i see a cross-fertilization between knowledge organization and
    TMs, not about TM concepts per se. murray mentioned jack park's upcoming
    book. if everything goes well, there will be a lengthy chapter by me
    on "topic maps in knowledge organization". but much more work is
    necessary.

2. May TMs suffer from more complex indexing models, like e.g. PRECIS did?
--------------------------------------------------------------------------
    rephrased: are more complex indexing models, expressed in TMs,
    more, or are they less effective than simpler indexing models,
    expressed in TMs? answer: it depends.
    this is a big, open research question, and one could say much about
    this. yes, they may suffer.

    in principle, the KISS (keep it simple, stupid) principle applies,
    i.e. keep it as simple as possible, but not simpler. if the requirements
    include to search for complicated concepts with complex interrelations,
    i see no other way. if not, it is a waste of resources.

    it is true that relational indexing never became very popular. e.g.
    the WRU system failed for several reasons, including that it was very
    costly, it was very difficult to reach consensus on which roles to
    assign, and it was awkward to use. but the problems of historical
    systems and that relational indexing is is seldomly used (one exception
    is BIOSIS), is _NO_ argument in principle against its potential.

    i interpret your question as follows: are TM proponents aware of
    work and failure with relational indexing, i.e. experiences about limits?
    answer: i am, but have not lost all hope. why?

    one might argue that the recent renaissance of relations in indexing
    (in AI, e.g. in conceptual graphs or for the Semantic Web), is due to
    a lack of historical knowledge of other branches of knowledge.
    in part, this is true. but in certain niches there is also the
    requirement for finer-grained semantical retrieval. where, e.g., you
    need to reliably search for process concepts (vorgangsbegriffe),
    you'll find nothing better than relational indicators (relationsindi-
    katoren) (cf. Fugmann).
    analytical thinking is in favour of relational indexing, where
    appropriate. i hold that there exist areas where the extra cost can
    pay off. this has to be determined in experimental settings.
    (i know of no retrieval test with TMs).
    there is also the hope that a richer ontology can better aid automated
    procedures employing this ontology, so it may even save money on mass
    data.

    to a certain extent i believe in the unique value of an intellectual,
    domain-oriented approach, complementing others: indexing is creating
    interpretations for (future) usage. it is necessary to analyze domains
    from the viewpoints of the discourse communities the indexing (the
    retrieval system) is targeted for. this will lead to insights about
    concepts and relations.

    it is true that relations introduce more chances for indexer
    "inconsistency". this is less a problem if we allow multiple
    interpretations (a "correct" indexing does not exist anyway, indexing
    theory/philosophy and epistemology tell).

    in sum: i think there exist application fields where relational
    indexing is of value, and implementing it with TMs, too. a sound
    knowledge about the state of the art in indexing and ontology
    engineering should methodologically guide against arbitrary linking.

3. Do domains exist where it may not be possible to adequately model
    their structure as TMs?
    ------------------------------------------------------------------
    my feeling is that it is possible to model any intellectual indexing
    with TMs (as they are hospitable enough). whether this is natural
    is a different question. so this is no limitation of TMs. rather,
    the question is about (1) can/should sophisticated indexing express
    arbitrary complex domain structures? (2) to which degree is this
    necessary/helpful for effective retrieval?

    murray mentions that believing in universals was the problem, and
    that viewpoints through scoping (a kind of partitioning the knowledge
    base) were one solution, also allowing for mapping between concepts.
    i agree. steve pepper's "towards a general theory of scope" is a
    valuable first step, but much more is to be done.

    in practice the question will be, if complex knowledge models will
    aid users in finding what they need, or rather confuse and distract.
    however, this is in part a question of the GUI (e.g. context filtering).
    a complex model does not necessarily mean complex navigation.

    i must admit that we don't have enough experience with sophisticated
    indexing of scientific matter in TMs yet. i know of no evaluation
    by library and information science or retrieval researchers.

    glenda, you did not provide an example of a potentially difficult
    domain. one might consider humanities and social sciences as more
    difficult to index than "hard sciences", but this is an open debate.
    i am currently working with viewpoints in TMs in a social sciences
    subdomain.

all the best
alex
--
----------------------------------------------
Alexander Sigel, M.A. sigel@...
Informationszentrum Sozialwissenschaften, R&D
Lennéstr. 30, D-53113 Bonn, Germany
+49 228 2281 170 tel, +49 228 2281 120 fax
Homepage: http://index.bonn.iz-soz.de/~sigel/

I have now put together new proposals for the language.xtm and
country.xtm PSI sets, and intend to submit these to TopicMaps.Org for
consideration in Montréal.

Feedback on these proposed new versions is very much wanted. My hope
is that we can discuss these now, so that we have progressed the
discussion and consensus as far as it can go before the Montréal
meeting. With luck the PSI sets may be adopted officially there.

The PSI sets are temporarily available from:
   <URL: http://www.ontopia.net/tmp/country.xtm >
   <URL: http://www.ontopia.net/tmp/language.xtm >

language.xtm differs from the previous proposal in the following ways:

  - a number of missing two-letter codes have been added, updating the
    PSI set with all changes made to ISO 639 up to May 2000

  - all languages that had only three-letter codes and no two-letter
    codes have been added. These now have PSIs of the form
      http://www.topicmaps.org/xtm/1.0/language.xtm#xxx

    This is sub-optimal, in that to create the correct PSIs from ISO
    639 codes, implementors must now know whether a two-letter code
    exists or not. This can be solved by using the additional PSIs now
    defined (see below).

  - all topics now have explicit subject indicators, so that importing
    the topic map from local file systems give the correct subject
    indicators

  - the topic for language-code has been replaced by a topic for
    language, so that there now exists a well-defined PSI for language

  - French names for all topics have been added (with some minor
    exceptions; contributions gratefully accepted)

  - additional PSIs and base names have been added for all the
    three-letter language codes (bibliographic and terminology)

  - the confusing xml:base attribute has been removed

  - the topics used for typing and scoping are no longer instances of
    the PSI "class"

  - the reference to the XTM DTD is now an absolute URI

  - the following codes have been changed
      - Indonesian: "in" to "id"
      - Hebrew: "iw" to "he"
      - Yiddish: "ji" to "yi"
    to reflect code changes made by ISO in 1989

  - the code for Serbo-croatian ("sh") has been removed, to reflect the
    fact that it was deprecated by ISO in February 2000 (one would
    assume the reason to be the existence of separate codes for Serbian
    and Croatian)


country.xtm differs from the previous proposal in the following ways:

  - all topics now have explicit subject indicators, so that importing
    the topic map from local file systems give the correct subject
    indicators

  - the topic for country-code has been replaced by a topic for
    country, so that there now exists a well-defined PSI for country

  - additional PSIs and base names have been added for all the
    three-letter country codes, as well as the numerical codes

  - the confusing xml:base attribute has been removed

  - the topics used for typing and scoping are no longer instances of
    the PSI "class"

  - the reference to the XTM DTD is now an absolute URI

No country PSIs have been added or removed.

--Lars M.

I must do a mediation after reading John Sowa's paper:

http://www.jfsowa.com/pubs/semnetw.htm

There is no question that professor Sowa is perhaps the leading scholar on
cognitive graphs and knowledge representation.

Dick Ballard and I have had the type of discussion that leads to
understanding something that is unexpected.

So I expect that a dialog between Sowa and Ballard will be constructive.
There are three forums involved.  The beadgame@yahoogroups.com is closed to
the public, since the Einstein Institute (U Conn) and the BCNGroup is
working on an experimental "bead game" technology.  Invitations to the
beadgame forum can be requested from beadmaster@...

The other two forums are the XTM forum xtm-wgt and the Communities of
Practice forum.  Both are totally open to public joining and participation.

The discussion will be of interest to xtm-wgt.

The community of practice community is not so deeply involved in the as yet
not yet fully developed knowledge representation theory and applications.
However, perhaps this distinguished forum might observe and make comments
about the issues that are raised in Sowa's paper and in the anticipated
responses from Dick Ballard and other knowledge scientists (Bo Newman, Art
Murray, Bob Shaw, and others).

Everyone, please be aware that these are the best minds in the world in
knowledge representation and that the issues are not yet resolved.

I can't believe that anyone would make the following statement
without joking or being incredibly naive:

> " As you might have surmised from my ontology paper, I have
> a single uniform coding system for all concepts. In any given knowledge
> base, absolutely everything (Content and Tool components) is characterized
> by a (Concept Sub-Code) and (Instance Sub-Code) pair."

That statement is at the level of claiming that C is a uniform coding
language for all knowledge because "absolutely everything (content
and tool components) can be characterized by a pair of pointers
to a (concept sub-code) and (instance sub-code) pair."

In 1957, Silvio Ceccato used the IBM 650 (a computer with a rotating
drum for memory) to represent all relations by a pair of pointers
and a code for the type of relation.  Conveniently, the 650 had
a word length of 10 decimal digits, of which the first two were the
code, and the next 8 were two pairs of 4-digit pointers.  But we have
come a long way from that "uniform representation".

If anyone is interested in a survey of the kinds of things that have
been done with semantic networks over the past 40+ years, I would
recommend my article:

    http://www.jfsowa.com/pubs/semnetw.htm

This article is a draft that will eventually be published in the
forthcoming _Encyclopedia of Cognitive Science_:

    http://www.macmillanonline.net/Science/ecs.htm

Following is the opening section.

John Sowa
__________________________________________________________________________

                            Semantic Networks

                               John F. Sowa

A semantic network or net is a graphic notation for representing
knowledge in patterns of interconnected nodes and arcs.  Computer
implementations of semantic networks were first developed for artificial
intelligence and machine translation, but earlier versions have long
been used in philosophy, psychology, and linguistics.

What is common to all semantic networks is a declarative graphic
representation that can be used either to represent knowledge or to
support automated systems for reasoning about knowledge.  Some versions
are highly informal, but other versions are formally defined systems of
logic.  Following are six of the most common kinds of semantic networks,
each of which is discussed in detail in one section of this article.

  1. Definitional networks emphasize the subtype or is-a relation between
     a concept type and a newly defined subtype.  The resulting network,
     also called a generalization or subsumption hierarchy, supports the
     rule of inheritance for copying properties defined for a supertype
     to all of its subtypes.  Since definitions are true by definition,
     the information in these networks is often assumed to be necessarily
     true.

  2. Assertional networks are designed to assert propositions.  Unlike
     definitional networks, the information in an assertional network is
     assumed to be contingently true, unless it is explicitly marked with
     a modal operator.  Some assertional netwoks have been proposed as
     models of the conceptual structures underlying natural language
     semantics.

  3. Implicational networks use implication as the primary relation for
     connecting nodes.  They may be used to represent patterns of
beliefs,
     causality, or inferences.

  4. Executable networks include some mechanism, such as marker passing
     or attached procedures, which can perform inferences, pass messages,
     or search for patterns and associations.

  5. Learning networks build or extend their representations by acquiring
     knowledge from examples.  The new knowledge may change the old
     network by adding and deleting nodes and arcs or by modifying
     numerical values, called weights, associated with the nodes and
     arcs.

  6. Hybrid networks combine two or more of the previous techniques,
     either in a single network or in separate, but closely interacting
     networks.

Some of the networks have been explicitly designed to implement
hypotheses about human cognitive mechanisms, while others have been
designed primarily for computer efficiency.  Sometimes, computational
reasons may lead to the same conclusions as psychological evidence.  The
distinction between definitional and assertional networks, for example,
has a close parallel to Tulving's (1972) distinction between semantic
memory and episodic memory.

Network notations and linear notations are both capable of expressing
equivalent information, but certain representational mechanisms are
better suited to one form or the other.  Since the boundary lines are
vague, it is impossible to give necessary and sufficient conditions that
include all semantic networks while excluding other systems that are not
usually called semantic networks.  Section 7 of this article discusses
the syntactic mechanisms used to express information in network
notations
and compares them to the corresponding mechanisms used in linear
notations.

This is a long "entangled bead"

for explanation see:

http://www.ontologystream.com/area1/primarybeads/bead1.htm

**
Protege (out of Stanford) is still the best hope of all publically known
knowledge technologies (In Our Opinion).


The topic maps has the quality of being a OSI standard but I do not think
that it has been able to focus on the true issues of knowledge
representation.  The following is a dialog between Drs. Paul Prueitt and
Dick Ballard.  This dialog is complex... but may reveal many or most of the
issues that are presently limiting the application of computer science to
the problem of knowledge representation and knowledge asset management.

****

Dr Ballard said:

"<..Paul> As you might have surmised from my ontology paper, I have
a single uniform coding system for all concepts. In any given knowledge
base, absolutely everything (Content and Tool components) is characterized
by a (Concept Sub-Code) and (Instance Sub-Code) pair."

<Dr Prueitt>

Richard, I do not have first hand experience in testing your encoding
structure.  Setting up such a test is precisely what OntologyStream is doing
for now several clients.  (I mean we are testing knowledge technology.)

I wonder if the business processes around you might consider enabling
ontologyStream and the US Einstein Institute (U. Conn) to make an evaluation
of the Mark 3?

In this way, we may be able to better communicate your innovations and a few
other innovations into the marketplace.

***

You also said:

"Mk 3 assumes layers where each layer may be constructed by different
knowledge base developers, so across these layers there may exist wide
differences in the identifiers attached each given model and even to the
instance assignments to a given model."

<Dr Prueitt>

This architecture fits perfectly with the stratified notion that each level
of several (many) layers taxonomy (ontology) should have gaps where the
measurement process must occur.

In the Intrusion architecture that Don Tobin and I are talking about, we
have four levels { data compression dictionary, Intrusion Detection System
output, Incidents, Goals(of attacker)/policy(of definder).

I have a test collection from one of my clients that is from the ACID
database (Army CERT Incident Database) - where CERT is a regional computer
emergence response team.

I need to develop a middle layer to the Tobin/Prueitt anticipatory intrusion
detection architecture (AIDA).

***

You also said (smile - never known you to talk so much)

"The integration of knowledge across different developers and
different copyright holders becomes a matter of relating the particular
(Layer, Concept, Instance) codes across all knowledge contributors, where
their concepts do in fact overlap."

<Dr Prueitt>

Now this is very exciting.  The BCNGroup Chapter directs us to mining
scientific publications for emerging intellectual property and claim this
property for the innovators and then allow (a choice) about whether this
property may be assigned to the BCNGroup for community purposes.  The
harvest of IP from the science community might be done with the Mark3 AND
the AIDA (modified).  The Value Proposition is huge, and Laramie is working
on capitalization of a process that will get the BCNGroup Membership
started:


http://www.fourthwavegroup.com/bcn/universal_sandbox_project.htm

  The "universal sandBox" is the key as the sandBox allows beadgames to be
played --> protected small virtual group collaboration at the innovation
level (example: the Int_group is now working out the Tobin/Prueitt AIDA IP).
In these protected virtual collaborations the development of IP is occuring
very rapidly while at the same time both a branding language (for public
facing communications) and a IP disclosure (for patent applications) is
automatically produced.

***

and also Dr Ballard said:

"The task of integrating layers and finding these concept overlaps in ways
that are not obscured by word choice, obscure definitional distinctions,
etc. is of course the ultimate problem of semantics and our primary use of
"content ontologies" (Kipfer, et.al.)"

<Dr Prueitt>

Dr Fiona Citkin is (in my opinion) the strongest mind in regard to the
notions of terminology comparison science (a concept that was and four other
Russian developed in the 1980s).  She and her husband Dr Alex Cikin are both
BCNGroup Founders.  Their work started in the Soviet Union, but has now been
placed within the protection of the BCNGroup in the United States.  (I say
this simply because of the great regard we (the BCNGroup Founding Committee)
has for the Citkin's work.)

***

and also:

"Now I do not know whether our cross layer semantic alignments and
integration is an intrusion to you?

"Again with respect to "syntagmatic" units <a, r, b>. In what way are your
concepts (a, b) different from r, i.e. is "r" a concept too, are concepts
recursive to all orders of logic. As you may have seen in the ontology
paper, our Primitive Sub-Codes are the only intrinsic property that we use
to distinguish the fundamental conceptual differences enumerated by our 18
(Ballard -Sowa - Peirce) ontological primitives. Mark 3 allows for even
greater range, our work says that Peirce's "thirdness" (Sowa's "Mediating"
concepts) is not enough, our work with "paths" readily flows through and
integrates Mediating Concepts so we think that whatever your highest order
"logical structures", "Paths" will always be at least one greater."

<Dr Prueitt>

yep...  *smile*,  the r are all from a set of stuff that becomes inference
rules when aggregated into a situational logic.  They are **completely**
separated from the concept atoms. They play the role of a (Peircean?)
semantic valance.  Pospelov told me (personal communication - Moscow 1997)
that there where 117 types of semantic valance and that this was language
independent.  Some of the thought that leads to this (perhaps now lost
Soviet research) is in Pospelov's translated but unpublished 1984 book
"Situational Control".   The notions that Finn developed establish a formal
foundation to situational logics that are open to new axiom reification -
and this has largely been lost also (except:
http://www.bcngroup.org/area3/pprueitt/kmbook/Chapter9.htm)

***

and also:

"Now by far the most interesting question about <a, r, b> is: are the bundle
of relationships you talk about based upon "r" ranging over a variety of
different concept  types (Concept Sub-Codes) or is your bundle a set of
relationship instances of the same type. That is the $64,000 question, as
they used to say.

"The question of relationship instancing is absolutely critical. Relational
databases define relationships by r type and by endpoint "identifier keys"
like "a" and "b". This is the absolutely fatal flaw in Codd's work and the
sure death ultimately of databases. Conceptual modeling formalisms like UML
that presume already the physical modeling choice and environment insist
upon the "endpoint" labeling of relationships and forbid relationship
instancing."

<Dr Prueitt>

Richard, there is no question in my mind that the set of r is a class of
types that can be organized into a period table.  (Why not?)  This wonderful
(and common sense) concept seems to have most clearly have arisen in the
classified work of the Soviet semioticians.. but was completely missed by
the Army (Tom Reader's group) and was not mentioned by reports on same
written by Peircean school Robert Burch (Texas Tech.)  But the wonderful
concept is all over Pospelov's "Situational Control", and in his numerious
presentations at the Army conferences 1994 - 1997.  My private conversations
on this confirmed that this wonderful concept was THE CORE to applied
Russian semiotics.

The idiotic attachment to bi-nary relationships, and rule based systems, and
the problems related to this artificial concept are what keeps us, as a
society, from developing knowledge technology...  The topic maps community
seems not to be able to go beyond this problem.  As for the death of Codd
normal form... <long live the new King, King XML/> we agree that the Codd
normal form is a instance of the Rosen category error.  Where topic maps to
become useful (as personal knowledge managemetn technology), it has to
better accommodate their distinction between addressable subjects and non
addressable subjects as a first step.

***

and also:

" Once you accept the "instanced vector" nature of even the binary
representations, then the conceptual jump to the "instanced vector n-ary"
relationship is an easy step particularly for physicists who teach routinely
of abstract phase space dimensions approaching Avogadro's number 6.023E+26
(cgs). These are, for scientists and engineers, routine conceptual exercises
although for real world problems the dimensions for most considerations
shrinks to 1.0E+3 and ontologies toward 1.0E+6."

<Dr Prueitt>

This is the notion behind "structural holonomy" where the natural linkage
between tokens are settled down onto (link or n-gram analysis).  This notion
has been seen to be formally defined as a mathematical tensor object (placed
in memory in a Forth OS) by a colleague of mine and a new notion of
relational database developed:

http://www.ontologystream.com/OS/PMIM.htm

I mapped his notion to the notions of Karl Pribram in holonomy theory of
brain function (Brian and Perception, 1991 Erlbaum).  The similarity of
these concepts needs to be traced using the M-CAM IP technology

www.m-cam.com

(They do link analysis on the Patent and Trade Marks database,and have some
wonderful visualization of the nearness of IP in IP spaces.)

***

  and also (the next six paragraphs)

"If we try to use an indexed database, to store n-aries it takes N(N-1)
index
entries to quickly capture and retrieve any n-ary and that complexity, for
real world problems, explodes to "non-computability". Hence, OLAP and any
number of retrieval solutions including mine. So why am I not seeing heavy
use of n-ary models?

"The n-ary vector is the natural representation of anything "conditional".
The larger the N the greater the number of conditions required to be
satisfied. One could choose aircraft design for example. List all the
different models, all the different configurations of airframe and engine,
all the different payload requirements, fuel capacities, all the performance
variables, altitude, maximum speed, range, etc. Each aircraft would be a
single vector instance and all the concept (instances) on that vector would
represent the particular design choices they made and the resulting
performance that they achieved.

"Some instances (top speed), (engine type) might be the same. Those concept
instances would show up on both vectors, but other choices and performances
would make the n-ary instances different overall. It is not important that
the number or order of concepts match up. They will be naturally different,
two airplanes may use completely different systems or technologies to
produce the same operational function.

"In the end this set of (r type = aircraft design - performance) vectors
describes all known aircraft design performance possibilities. Now the
reason that a particular aircraft engine produces a given top speed is not
explained by this vector. That relationship is explained deeper in the
knowledge base at finer levels of conceptual granularity. What the n-ary
focuses us on is the possible decisions and their outcomes. To engineers,
this is the "trade space" where on the basis of existing designs and
technology, the constraints add up to require specific choices and
possibilities. They can see in moments the experimental consequence of
trading one choice off against another while focusing on the most demanding
performance requirements.

"Obviously designers may hope that a different design or technology might
let
them do something new. Knowledge of the trade space is one of the key
elements of expertise and knowledge superiority, the size of the base and
the speed in using it. In knowledge management, knowledge of trade space is
among the first to be lost (requirements is another)."

*

<Dr Prueitt>

yep.  and this is the basis for knowledge assets management.  It is very
simple.

***

(and also: the next six paragraphs.)


"Now we come to the issue of procedure versus declarative knowledge. It
takes
years of engineering, and billions of dollars in labor and computing to
produce one such vector result. What if at night someone had explored that
design space by running all the other reasonable design choices through the
simulation or performance estimating software and saved all those other
vectors. We call this "sampling the design space" or "learning". It can
happen once long before we need it again.

"Years later the procedure and those who know enough to make it run may or
may not be around. Today it takes up to 14 years to produce a new airplane,
so the odds of exploring that space later is pretty remote.

"The basic problem with most design problems is that design is a one way
process, propose a design, then spend years testing and perfecting it.
Process and procedure is too often a non-reversible activity.

"By contrast, the customer does not want design. What they are looking for
is
"something", maybe anything, with a given performance. Stored as a
"declarative vector" the aircraft design - performance space has no
preferred direction. It is just as easy to start with performance and see
all the designs that come closest to that required and all the other,
incidental characteristics. Now what is the computing cost for examining the
stored declarative trade space. By knowledge theory at worst it should be
proportional to information content and virtually instantaneous.

"What would be the cost instead of browsing that space with the design
(programs) procedure? It should take less than 14 years, if we have anything
left to start with.

"The declarative knowledge representation is about learning once and never
forgetting. The procedural approach is about never learning and always
starting over. The procedural approach makes sense in an information rich
environment where things are never the same or where the cost of remembering
is the greater burden. The declarative approach makes sense in a meta rich
world were knowledge is stored in n-aries and put in places that never die."

*******

<Bead master/>

Thank you, Dr Prueitt and Ballard, for the presentation of concepts.  These
have been forwarded into the beadgames (protected sandboxes), beadgame and
Int_group for further discussion.  Anyone wishing to join these two games
may apply by sending a message to me at

beadmaster@...

With respects and appologies to everyone.

Nikita, Piotr,

I agree that scope is a mighty yet heavily overloaded concept. And like all
sophisticated 'tools' it can either be of great use when it is used in a
careful and consistent manner, but hazardous and confusing when used without
any special care and methodology. Like many other aspects, I propose, the
correct use of scope is part of a well planned topic map design process. A
TM design will absolutely be based on strict type hierarchies and schemata.
In such a scenario scope can also be represented, as Piotr said, by the
superclass hiearchy of a topic. If you understand context as the environment
or domain that sorrounds or includes the 'topic of interest' (which is a
common definition in literature), context can even be interpreted as the sum
of all associations and associated topics, i.e. the complete topic map from
the point of view of the topic of interest.

This may sound rather academic, but it illustrates my point: there are many
ways to 'model' or represent context, and scope is just one means which is
explicitly declared in the topic map language. So I think scope should be
used as a syntactic shorthand (note that the <scope> element is again just a
shorthand for a topic association expressing scope, where one member plays
the role of being scoped and the other playing the role of representing the
scope) in cases where the more elaborate modelling of context by other means
is either too tedious or may be beyond the intended 'expressiveness' of the
topic map. In this way it might well serve as a mechanism to 'tag' topics
for interpretation by application systems for expressing specific user
contexts (access level, user profile etc.) Or also for a quick way for
expressing context without a complete redesign of the topic map design.

In other words: I guess the current very interesting discussion around the
meaning and use of scope will become clearer once topic maps have been
understood by the community to require a similar design approch like for
instance OO modelling.

concerning the ideological debate about whether a topic itself can have
scope - or only topic characteristics should be scopable: I support the 2nd
opinion. I have to admit I havent yet understood what you can gain from
allowing scope to be attached to the topic itself. Maybe someone could
explain this to me. On the other hand I am afraid changing the XTM spec in
this respect might cause a lot of work as it will also affect the topic map
merging rules.


Heiko


------------------------------------------------------------
Dr. Heiko Beier        -  CEO               - moresophy GmbH
Brienner Str. 54b      -  80333 Munich
Tel: +49-89-523041-71  -  Fax: +49-89-523041-89
------------------------------------------------------------
          moresophy -> more sophy -> think meta
------------------------------------------------------------

| -----Original Message-----
| From: Nikita Ogievetsky [mailto:nogievet@...]
| Sent: Tuesday, July 24, 2001 5:50 PM
| To: xtm-wg@yahoogroups.com
| Subject: Re: [xtm-wg] re topic.scope
|
|
| Piotr,
|
| Scope tends to be very overloaded.
| And this is the source of confusion.
| Its main purpose is  to assert a context within which a certain
| statement (association) holds.
| Then people start punting all convenience shortcuts into it and get
| confused.
|
| Topic can not "make sense in a certain context"
| Topic by itself does not make any sense AT ALL!
| Topic's characteristics do.
| And YES you can scope all topic characteristics.
| Lets be clear on this.
|
| As Steve P. says,.
| >Themes specified on a <topic> element are only "inherited" by names
| >and occurrences that are subelements of that element.
|
| So in ISO if you put a <scope> on a <topic> level it will mean that
| all topic characteristics defined by this <topic> element
| "make sense in a certain scope".
|
| But stop, remember that there are might be multiple
| <topic> elements for a single topic (subject) plus associations ...
|
| However, I found it a good practice to use
| typing for hiding topics from certain user groups, for example.
| (similar to Piotr's and Ivan's suggestion, I guess)
|
| --Nikita,
|
| ----------------------------------------------------------
| Nikita Ogievetsky               Cogitech Inc
| XML/XSLT/XLink/TopicMaps   Consultant
| nogievet@...   --   (917) 406-8734
| http://www.cogx.com     Cogito Ergo XML
|
|
|
|
| ----- Original Message -----
| From: "Piotr Kaminski" <pkaminsk@...>
| To: <xtm-wg@yahoogroups.com>
| Sent: Tuesday, July 24, 2001 6:17 AM
| Subject: Re: [xtm-wg] re topic.scope
|
|
| > I have a potentially silly suggestion to make:
| >
| > For topics, does type classification not serve the same purpose as Ivan
| > would like scopes to serve?
| >
| > So if a topic only makes sense in a certain "context", just make that
| > context into a class, and make sure the topic is of that class,
| and others
| > that don't belong in the context aren't.  It would be the class of "all
| > things that make sense in <fill-in-the-blank> context>.
| >
| > The obvious problem with this interpretation is that the topics' classes
| > will not match the associations' scopes (unless a topic can be both a
| class
| > and a scope simultaneously? hmm...).  Which of course brings up another
| > suggestion:  why have separate concepts of class and scope for
| associations?
| > Just allow them to have multiple types, and use these for
| scoping.  What's
| > the big difference between classes and scopes that would prevent this?
| >
| > Steve P.:  I have not yet read your paper on scoping, though it's on my
| > immediate to-do list.  If the answers to my questions are all contained
| > therein, don't waste your time answering my rantings.  :-)
| >
| >         -- P.
| >
| > --
| >   Piotr Kaminski <pkaminsk@...>  http://www.csr.uvic.ca/~pkaminsk
| >   "It's the heart afraid of breaking that never learns to dance."
| >
| >
| >
| >
| > To Post a message, send it to:   xtm-wg@yahooGroups.com
| >
| > To Unsubscribe, send a blank message to:
| xtm-wg-unsubscribe@yahooGroups.com
| >
| > Your use of Yahoo! Groups is subject to
| http://docs.yahoo.com/info/terms/
| >
| >
| >
| >
|
|
|
| To Post a message, send it to:   xtm-wg@yahooGroups.com
|
| To Unsubscribe, send a blank message to:
| xtm-wg-unsubscribe@yahooGroups.com
|
| Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
|
|

Murray Altheim wrote:
> ... snip ...
>
> Robert,
>
> I completely understand the desire to create a simplified authoring
> syntax, absent a good GUI authoring tool.

I don't know how you define 'good' but empolis k42 provides
already since its version 1.0 (released January 2001) a Web
Authoring GUI and k42 version 1.1 comes with an even improved
Web Authoring GUI. With the XTM import/export of k42 you
have an XTM GUI authoring tool.

If you want to figure out if k42's GUI is 'good' get a free download
at http://k42.empolis.co.uk (full functional, only restricted
in the size of the map).

> ... snip ...

Cheers,
--Holger

--
Dr. H. Holger Rath <holger.rath@...>
Director Research & Development
empolis GmbH, Havelstr. 9, 64295 Darmstadt, Germany
http://www.empolis.com/ -- mobile: +49.172.66.90.427
phone: +49.6151.380.292 -- fax: +49.6151.380.488

* Robert Barta
|
| As Lars was concerned adopting some approaches into LTM 1.x,

My concern was backwards compatibility. Lots of people know LTM
already, there are LTM topic maps already, there is LTM software
already, ...

| I thought it worth to give it an independent try. I was discussing
| with Lars this issue extensively and our understanding was - Lars
| correct me in case - that there is room for another language and
| time/support will tell.

Well, like I told you I would have been happier had you not done this,
but like I also said I don't imagine that AsTMa will any harm.

You should write some BNF, though. It would make implementation easier
for you, and it would also help interoperability if other people
implement AsTMa.

--Lars M.

* Murray Altheim
|
| I currently use LTM 1.0, with a 1.1 version looming. (I've been
| discussing changes with Lars Marius, and am hoping to discuss this
| further with him in Montreal)

I would have loved to discuss LTM with you in Montreal, but
unfortunately I am not going there. So I'm afraid we'll have to
discuss this via email until Orlando. (Unless you are going to
Copenhagen. :-)

--Lars M.