On Saturday 13 November 2004 17:09, Kurt Conrad wrote:
> Finally, today's brain teaser for all who care to respond is:
>
> What is Semantic Harmonization? (01)
Some definitions: (02)
SYSTEMS INTEGRATION = Enabling the communications that allow systems to work
jointly toward a goal. (03)
[That is roughly how we defined systems integration in a recent project at
NIST. This definition only applies to software systems and only concerns a
part of systems engineering.] (04)
SEMANTIC INTEGRATION = That part of systems integration that concerns whether
or not messages, correctly received and decomposed, serve to direct the
recipient to perform the desired behavior. (05)
SEMANTIC HARMONIZATION = misplaced emphasis ;^) (06)
* * * (07)
I suppose that needs some explanation. First, I think the goal for most of us
is to get information systems to serve business goals as effectively as
possible. Because thing change (new business partners, new business
processes, new products etc.) an important part of this goal is getting the
components (parts of a system, or new partners) to integrate as effectively
as possible. (Here "effectively" = some balance of cheaply, quickly, and
reliably). So the principle goal includes doing systems integration as
effectively as possible. (08)
Two more points for background: (09)
(1) The path of progress in systems integration has squeezed a good deal of
waste out of the job of integrating systems. That 'waste' was in the form of
reiterating the engineering details that needed to be resolved. Far less
progress has been made in resolving the issues surrounding semantic
integration. (010)
(2) The emphasis in the terms "semantic integration" and "semantic
harmonization" ought to center around "desired behavior." This is because, as
far as we are concerned as the engineers of systems, communicating meaning
comes down to getting the right behavior out of the recipient of the
communication. For the engineers of systems, semantics is about behavior. (011)
With that said, I'll try to put a little more work into a definition of
semantic harmonization: Semantic harmonization, I suppose, is the task of
eliminating unnecessary differences in the content of messages. I suppose
further that the "content" is "semantic content" -- it concerns distinctions
that reflect differing intended behaviors on the part of the recipient. But
then, if distinct behaviors were intended, why would we want to eliminate
them? Weren't those distinct behaviors a business necessity? (012)
I'll try again: (013)
Semantic harmonization is the alignment of terms so as to share a common
understanding of the universe of discourse. (014)
But what do I mean by "*the* universe of discourse?" The more I say about a
subject, the more I define its boundary, and the more likely it is that you
will see things from a different perspective. And sooner or later, something
I say about the chemical industry won't apply to the petrol-chemical
industry, much less banking. (015)
To address this problem, we have invented a notion of "contexts," the purpose
of which is to partition the universe of discourse into "smaller universes of
discourse," wherever inconsistencies in the use of terms arises. It would be
nice if contexts decomposed in hierarchical fashion, but I don't think this
is the case. Nor do I think that we can enforce some order on the
proliferation of concepts by applying a classification scheme. (e.g. the 8
dimensions of context described in the CCTS). That might work for a while,
but as the number of entries increase, the distinctions become more nuanced.
As a systems integrator, I'll need to know exactly what those distinctions
are. An informal natural language description of the distinction might not
tell me, because it may assume that I have understood something about your
domain that isn't explicit in the definition of the isolated terms of the
message. (016)
Notice then, that as a systems integrator, I have to do some of the same work
of understanding your domain by means of the registry as I'd do by the old
way -- that is, by studying a conceptualization of your information systems.
And in both methods I still have to be very familiar with the business
processes of both communicating parties. (017)
We have moved the system integrator's task of understanding the meaning of
terms from one of understanding the conceptualization of the information
system, to one of understanding the distinctions intended in the various
registry entries. Nonetheless, that might be progress, since now I am at
least assured that there is *some* documentation. (018)
In a registry-base approach, semantic harmonization involves a software tool
(the registry). Thinking more generally about tools for semantic
harmonization, what might they provide? What if the tool allowed us to
explicitly define the distinction intended in a formal language, rather than
just classify it and explain it in natural language? The 'formal
language' (e.g. KIF, OWL) would include a standardize vocabulary in the form
of an upper ontology that is common to the largest universe of discourse. In
that arrangement the systems integrator might be able to use the tool to
identify the relevant business distinctions itself, and help validate the
message types constructed. (019)
* * * (020)
I still like the first definition, "misplaced emphasis." But, if you find none
of my definitions of semantic harmonization above satisfying, I wouldn't
blame you. (021)
I'll take this parting shot: maybe while we *think* we are "harmonizing the
semantics" of our business communications, what we are actually doing is
harmonizing the business processes themselves; that is, eliminating
unnecessary differences in the way we do business so as to share common
practices. When we do that, we are relieved of the need to communicate
peculiar circumstances of practice. (022)
How far can the harmonization of business processes go? In this community, the
opening statement of many presentations is something like "Does the
pharmaceutical industry really need a different purchase order than the
medical community?" The speaker always assumes the answer is "no" but I am
not so sure. After all, business processes vary because we are doing various
thing! And I'd bet that most of us are not sufficiently well-versed in
multiple industries to confidently answer the question. (023)
--
Best Regards,
- Peter (024)
Peter Denno
National Institute of Standards and Technology,
Manufacturing System Integration Division,
100 Bureau Drive, Mail Stop 8260 Tel: +1 301-975-3595
Gaithersburg, MD, USA 20899-8260 FAX: +1 301-975-4694
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Unsubscribe/Config:
http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (025)
|