[topicmapmail] Summary to date of "context" and "bottom up"

[Jack Park]

I have name-tagged all the posts in this series for the following reason:
the posts illustrate a collaborative process in progress.  Along the way, we
see chuncks of agreement, and chunks of disagreement.  Where we need to go
from here is toward convergence.  IMHO, we do not yet have enough body mass
involved to get there.  I am wondering if we shouldn't copy this post into
other groups, e.g. Standard Upper Ontology, the PORT list, and others, as a
means of opening up the meme pool.  For myself, I am beginning to imagine an
ontology growing out of this thread, one that encodes all arguements, and
that enables the search for convergence.
 
Originally, I actually split up posts. Later in the process, I allowed name
tags to take, within their body, other name tags; that to more accurately
reflect the specifics of injected ideas.
The actors are:
<BV> Bernard Vatant
<MB> Martin Bryan
<JDM> James David Mason
<JP>  Jack Park (if I ever jump in).
I would like to see others enter this thread.
************************************
<BV> The debate here is very rich and multidimensional, and I almost forgot
why 
I joined in !
 
So I'll try to stick to it here : my main concern, in the frame of the
project 
presented previously (cooReso), is to know if topic maps standard has the 
capacity to comply with the following features. I'm quite convinced the
answer 
is "Yes" for the first point, I'd like to be convinced that it's "Yes" also
for 
the second one, and maybe you'll help me to clear that out.
 
1. A community of contributors has to manage a context : define its
ontology, 
scope, subjects, topics, names, associations, types, etc ... and build the 
occurences data base. 
</BV> 
<MB> Not necessarily "A community" - its often a set 
of communities, as the current ebXML work for electronic business semantics 
illustrates clearly. 
</MB> 
<BV> Thank you for your different viewpoint. Anyway, maybe we have more
"definition problems" , coming for rather different contexts and
communities, than fundamental opposite views. Hope you won't consider the
following setting of "definition problems" like "hair-plitting" :o)
<definition problem>
As soon as people have to share their ontology, aren't they de facto a
community ... even if they don't feel like it to begin with? In maths, a
set is defined by the sharing of a common characteritic property, does that
definition hold there? It's true that the term "community" is widely used
now and maybe in a too fuzzy way. Since we are you and me there trying to
define things, do you feel like we are in the same community? There is kind
of bootstrapping definition : community is defined by the context they
share, context is defined by community sharing it.
</definition problem>
<MB> Let me use your own community to disprove the concept of it being
singular. 
You are a mathmetician - correct? Now my brother was trained as a
statistician, 
and a friend of our is currently considering being trained as an actuary.
Now 
consider what the term "number" means to yourself and to my two 
"mathematicians". (Or try to determine what set means to the three of you!)
They 
share a common term, they do not share a common definition, but they do
share a 
set of processes that can be applied in certain contexts to these objects.
They 
also have processes specific to their disciplines that the associate with
the 
same term.
 
Now let me take you to another sort of community. The set of processes
involved 
in sending goods from one country to another. The involved parties are 
Manufacturer, Stockist, Shipper, Customs, Customer. They may all have a 
different terminology for the same concept: perhaps called BatchSize, 
PurchasableUnit, PackedQuantity, ShippedQuantity and QuantityOrdered. They
can 
agree on the relationships between them but not on a common ontology Yet
there 
has to be a mechanism for sequentially converting from one to another.
 
What we need to do is to define a sharable set of common objects with known 
associable processes. (The process of identifyin the commonality they share
is 
fundamentally the difference between a theory and a practice.) To do this we

need to divorce local nomenclature from ontological nomenclature, and local 
processes from commonly shared processes. We need to know that this set of 
mathematical operations are performed in the context of actuarial science,
this 
set in the context of statistical analysis, and this set in the context of
set 
theory.
</MB>
</BV>
 
<BV> I precise what I mean here by "context " : a 
field of knowledge, or interest, or activity, or experience (whatever) 
pragmatically bound by the community of people sharing it and speaking a
common 
language, meaning by that the semantics are non-ambiguous in the community. 
</BV> 
<MB> I have a different view of context. To me it is the way of 
identifying which subunits of the set of communities that built the
onotology 
use the topic for which processes. Most communities use a topic for more
than 
one process, but to understand accurately what they mean you have to
understand 
what process they are associating the term with at each of its occurrences. 
</MB> 
<BV>
<definition problem>
I agree that your definition is surely more experience-grounded and more
accurate and less naive than mine.
Ok to refine the context as follows : "field in which a set of people (call
it a community or not) use the same ontology to deal with the same
process". But it does not change the main point I'm up to: you can't define
context independently of people involved in it.
</definition problem>
</BV>
<MB>
>From the above business related argument I would claim the following to be
more 
accurate: "The field in which they can agree on the processes that can be 
applied to a set of named terms that share a common definition". (Its the 
definition that counts, not the name that is applied to it.) 
</MB>
<BV> That's of course an idealistic viewpoint, since in detail, semantics 
are different for each of us, as Jack pointed out. But if you don't split
hairs, 
this definition is an operational one. Anyway I insist on this point : a
context 
is "embodied" in a community. And it's up to this community to build and
clarify 
its ontology. It's an interesting social work in itself. That's what I meant

earlier in my answer to Martin (or maybe Jim ?) to be a bottom-up approach :
the 
community and its ontology, whatever its size, is the "local" or "ground" or

"bottom" level, and if we have to build more general ontologies, we'll have
to 
merge them somehow. It has nothing to do with the way you organize things
inside 
context. Sure if you have to deal with car pieces, you've better have some
neat 
hierachy such as trademark/model/year/type/engine/... or the one Martin
quoted 
about air travel reservations ... 
</BV> 
<MB> But if you are an engine 
manufacturer it may be more important to define
engine/type/model/year/trademark 
if you want to know that your V16 engine was used in the 2000 version of the

Ventura that General Motors issued under its Vauxhall trademark. (Context is

very much dependent on the user of the information, not on its creator.) 
</MB> 
<BV> I can't imagine how it could be the other way round : local communities

refering to some "general ontology" and trying to adapt it to their context.

Which means we may have local top-down approach, inside a local context, but
not 
in a "global context" if that expression makes sense. My opinion is that it
does 
not make sense anyway, and the arguments for that are both technical and 
political. 
</BV> 
<MB> Note that the context does not necessarily change the 
underlying ontology, but can change the order in which items in the ontology
are 
considered relevant. 
</MB> 
<BV>
<definition problem>
"The order in which items are considered relevant"
Suppose you mean by that the association architecture.
I thought that was considered internal to ontology : what you mean by one
term in one context is linked to what you mean by associated terms, and the
way you associate them. What I understand is changing context does not
necessarily change the vocabulary, but the extension of terms semantics and
their eventual hierarchy, as you point out very accurately with engines and
cars.
</definition problem>
</BV>
<MB> Not sure what you mean by the term "association architecture". What is 
important to me is that there are multiple possible routes to reach the same

concept, and that on some of those routes the same set of ancestors can
appear 
in different orders. I consider this to be external to the ontology because
of 
the possibility of reordering, which is not inherent in most ontologies.
</MB>
<BV> The technical view is that whoever sets the 
"upper level", they are themselves a community (see SUO), and what they will

call "general" is only what they suppose at the moment to be general in
their 
implicite community ontology, which is *always* cultural-context biased, and

will most certainly be irrelevant elsewhere. The political view is that
their is 
no universal concept whatsoever, and trying to setting one is de facto
trying to 
impose its own concept as universal. That's what I call "unique thought".
It's 
an old trap we must always be aware of, because it's very temptative. Maybe
I'm 
myself falling in it right now, you'll tell me. 
</BV> 
<MB> But for global trading you need a "universally accepted terminology",
with 
multiple names but one (translatable) definition. 
</MB> 
<BV>
<definition problem>
Of course, but global trading is itself a context ;-)
 <MB>
 Yes, but it is itself a set of processes, as shown above, 
 each of which is its own context.
 </MB>
Don't get me wrong. From a technical viewpoint, standard terminology is
wonderful and I'm always amazed by the tremendous amount of work done in
this field, permitting to communicate as we do today. What I am up to is
more in the "education-information-opinion context" than in the
"trade-business context". 
 <MB>
 Even in a information oriented field such as education or politics (which
is 
 just a matter of opinion!) the context in which you use the term can affect

 its perceived meaning. For instance the term "mathematics teacher" has a 
 different perception to someone in the Ministry of Education from that of 
 someone in the Treasury :-) The processes involved are different in each 
 case. Try asking someone in each of these departments to draw up an
ontology 
 that applies to mathematics teachers. Maybe they would come up with:
 
 Education/Secondary/Sciences/Mathmatics/Teacher and 
 PublicEmployee/PrefecturalAppointment/Level13/Teaching/Mathematics
 
 Same object, same meaning of end term, different contexts in which to find
the 
 object/meaning. Totally different ontologies ;-(
 </MB>
Saying there is no universal set of concepts
through which analyze and organize all our thoughts and "views of the
world" is not necessarily in contradiction with a global agreement on
technical standards to communicate with.
</definition problem>
 <MB>
 Ah, now we come to the real key - how do we communicate the relationships 
 between objects/things/occurrrences/classes/concepts/topics and the
contexts 
 in which they are used. Topic maps allow you to identify concepts and
topics 
 and differentiate them from the occurrences of things that represent them. 
 They also allow for the identification of the relationships between them in

 a way that allows different sets of associations to be assigned to
different 
 user communities. What they do not allow us to do is to identify the 
 sequence of associations that leads to the need for a particular topic. 
 (They have no concept of class/subclass or of precedence.)
 </MB>
One of my main concerns in that debate is how to find safeguards against
technical standards imposing "thought standards". We know pretty well that
our use of specific technical tools shapes in some way our "view of the
world". Like it or not, my daily use of Microsoft products shapes somehow
the way I think, and puts me in the "MS users community" :(   ...  which
does not mean I share all of Bill Gates ontology (I've good reasons to
think it anyway)
</BV>
<MB>
We are all shaped by the thought standards (or, in the case of newspapers,
"lack 
of standards") that surround us. The trick is to divorce the thought
patterns 
from the information patterns. Lateral thinking is key here. It is
postulated on 
the idea that there is no known association to start with, but that applying
a 
known association to a currently unconnected pair of known concepts pays 
dividends. So given two topics, which set of the associations we have agreed
are 
useful to our community can be applied to connect these two topics?
 
Context sits above ontology, not within it. It acts as a filter to the
ontology. 
It ca be used used to select topics based on their scopes rather than their 
ancestory. It can also be used to identify facets that provide user
dependent 
views on a topic map. Thats why <number scope="statistics"> and <number 
scope="arithmetic"> will continue to be needed for occurrences in topic maps

about numerology :-)
</MB>
<BV>
 
2. We have to merge contexts while respecting their local ontologies.
 
That's the very tricky point. I think the solution is there again only
local, 
which means it's up to a "context community" to define why and how its
ontology 
is to be linked to neighboring ones. I mean by neighboring that they share
some 
topics, some associations, some occurences, but not all of them, and that
their 
scopes are somehow overlapping. Neighboring contexts should be merged
through 
what they share, and keep autonomy for what they don't. This "neighboring" 
relation is to be used there again on a very pragmatic ground. To take again
the 
example of chemists and baseball players, it's up to them to decide if they 
declare neighboring through the problem of ball engineering, or if it's a
too 
farfetched association. (I'm too ignorant of baseball myself, not speaking
of 
chemistry, to decide !)
 
The neighboring relation is not transitive, which prevents us from any
global 
merging. Some will consider it like a major shortcoming. Considering what I
said 
earlier, you'll understand I'll take it as a precious safeguard against any 
global temptation.
 
So I set again the question : can topic maps standard comply to #1 and #2 
exigences ?
</BV>
<JDM>
I generally second what Martin (<MB>)has said. 
 
My experience has been that it's important to decide what sort of thing
you're looking for in any taxonomy. "trademark/model/year/type/engine/"
works if your emphasis is on what options are available at a given time for
a particular model, but it isn't a good model for other things. For example,
Volkswagen makes a 1.8L turbo that they put in half a dozen different models
of Audi and VW products. if you're coming into the subject with an interest
on engines, "trademark/model/year/type/engine/" doesn't work, but Martin's
reversed order does work. 
 
We've found that hierarchies are useful for some contexts, and that it's
important to keep hierarchies structured by consistent principles (i.e.,
keep "isa" hierarchies separate from "isapartof" hierarchies). Most of our
projects have been in various fields of manufaturing, and we find that we
have to have quite a few hierarchies. Some are common across applications-we
use subsets of materials properties hierarchies in numerous places. We find
component breakdowns to be a common type for manufactured products, but for
those we can't just subset some generic collection the way we can for
properties of alloy steel. But we find that our topic maps link into those
trees in many different ways. We haven't done enough projects yet to have a
good sense of what all the kinds of TM links will be.
</JDM>


============================================================================
This message is intended only for the use of the Addressee(s) and may
contain information that is PRIVILEGED and CONFIDENTIAL. If you are not
the intended recipient, dissemination of this communication is prohibited.
If you have received this communication in error, please erase all copies
of the message and its attachments and notify postmaster@verticalnet.com
immediately.
============================================================================