At 2003-01-20 03:48 -0500, William E. McCarthy wrote:
>In my mind, an ontologically-oriented team
>would feel incomplete if they had to walk away knowing
>that a significant component of this knowledge was left
>unspecified in their efforts (for example, "where
>approximately are we in this deal?"). Much of it was
>certainly embedded in the paper documents of years past,
>but simply stopping there would be (again in my mind) an
>incomplete effort no matter how exhaustive their levels of
>effort were at unearthing document-bound knowledge. (01)
In light of the top-down/bottom-up/mixed models previously discussed,
I'm not entirely convinced that there is any one "right" or "best"
approach. (02)
Given that most of these ontological efforts (whether or not they are
recognized as such) take place in a business setting, every attempt
at specification will have to balance short-term needs, long-term
opportunities, and cost. The optimal balancing point for these
potentially competing interests not only varies with business context
but across time. (03)
Likewise, the optimal development method can be expected to vary
across applications. The property of being "incomplete" is often the
driver of business value, especially for behaviors dependent on
event-specific knowledge. (04)
It's really a question of sufficiency, not completeness. Sometimes,
we need top-down alignment around intent (and can fill in the
low-level details). Sometimes, we need to align around the low-level
communication protocols and will provide our own context. Sometimes,
both need to be specified. (05)
But Bill raises an important and timely issue for this group and its
informal relationship to the UBL effort. Namely: How the UBL effort
could apply ontological engineering principles to improve their
effectiveness as a committee and the quality of their work product. (06)
Toward this end, let me pass on my understanding that the UBL effort
was launched in response to the top-down (process modeling-based)
approach of the ebXML initiative. It was not intended to undermine
or deny the value of top-down approaches for standardizing business
documents. (07)
At the same time, top-down approaches can be cumbersome and quite
time-consuming. In the case of a core set of business documents
(purchase orders, invoices, etc.), the basic structural models for
these documents have already been agreed to, by
convention. Formalizing these conventions using XML schema is
expected to reduce the cost, time, and effort necessary to transmit
these documents electronically. In this scenario, the processing
models remain largely outside of the scope of the formalization
process, just as the do today. (08)
Likewise, industry-specific requirements and other customizations are
expected to be handled through an extension mechanism. UBL, itself,
is expected to represent an 80% solution. (09)
Even accepting Gruber's definition, how do we scope
"conceptualization"? Does it have to include some (or all) of the
potential behavioral implications, or can it be limited to a specific
class of knowledge artifact? (010)
Do we believe that specifying/formalizing an artifact independent of
process is inherently problematic? If so, is this expected to drive
critical problems that preempt utility or merely that their would be
comparative advantages to specifying a more complete conceptualization? (011)
How do you know if you need more "top-down" or "bottom-up"? Does it
matter whether you do the "top-down" or "bottom-up" first? (012)
Perhaps even more fundamentally, are we all working with the same
definitions of top-down, bottom-up, and mixed? (013)
Speaking from my personal experience with the UBL committee, I've
seen them struggle with the optimal balance between a focus on the
structural relationships and the various behavioral contexts
(analysis, design, and eventual utilization). The need to focus more
robustly on the utilization scenarios will become even more important
in the future, as the team starts to deal with the customization
model (under the rubric of "Context Methodology"). (014)
/s/ kwc 2003.01.21 12:28 (015)
___________________________________________________________________
Kurt Conrad
2994 Salem Dr. 408-247-0454
Santa Clara, CA 95051-5502 408-247-0457 (data/fax)
http://www.SagebrushGroup.com mailto:conrad@xxxxxxxxxxxxxxxxxx (016)
--
To post messages mailto:ontolog@xxxxxxxxxxxxxxx
An archive of the [ontolog] forum can be found
at http://ontolog.cim3.org/forums/ontolog (017)
|