Thursday, November 24, 2005
Center of Excellence Proposal
à
ONTAC stands for Ontology and Taxonomy Coordinating Working Group
It is a working group of
Semantic Interoperability Community of Practice (SICoP)
Discussion between Paul Prueitt and
Matthew West
Editorial note: Matthew made responses that are indented and in large type
I
liked what I saw in the ISO 15926 Part 2, where there is a general
ontology.
http://www.infowebml.ws/ECM4.5/ECM4.5.html
MW: You have been busy to find this website. The ECM
here refers to the EPISTLE Core Model, which is the name under which the model
was developed. This version purports to
be ECM 4.5. The version standardized as ISO 15926-2 was in fact ECM 4.5.1, so
there will be some differences. The previous link
you have used to SC4Online is authoritative.
It
also has the flavor of an object oriented class hierarchy, and thus has the
usefulness that so many programmers have become aware of. Inheritance, encapsulation,
I
have a question for those here.
Why
is ISO 15926 not sufficient as a general ontology for web services across all
"business exchanges". Has
the generality of this set of 201 entities been tested in other domains - not
oil and gas.
MW: As I mentioned elsewhere, it is being used by Shell to model its Downstream Business, so financial transactions, customer relationships, geographic objects. This use is a good distance from refinery and offshore equipment. It was always intended that it should be suitable more widely, but a clear focus is also helpful in developing an upper ontology.
Is
there a move towards ending the standard setting process and actually using
this specific set of "entities" in oil and gas?
MW: They (or their predecessors) have been in use for
the last 10 years, especially in the Norwegian Offshore industry. However,
there is no end to the standardization in sight. Part 2 is considered
relatively stable, but there is a great deal of effort currently going on at
the next level down in the Oil and Gas equipment domain so that designs of offshore oil rigs (with a
value usually in excess of $2b) can be handed over from contractors to
owner/operators. We need a common language for things like pumps and valves, as
well as the more abstract concepts in ISO 15926-2. These are being standardized
in Part 3, and an ISO Register is proposed to hold these and to allow extension
of them (not necessarily restricted to Oil and Gas).
As
long as the standards are not stable, the businesses that benefit are limited
to those few that are developing standards.
The larger (theoretical) value from the notion of Semantic Web (that
part which is correct) is not accessable, if the bulk of the standards are not
set.
MW: I agree. Dissemination takes a while (years) and
people are naturally reluctant to commit to something they think will change.
On the other hand, our experience has been that those that did take the plunge
found significant benefits from doing so.
If
there was an ontology hub (freely available sets of entities such as ISO 15926)
where each industry adopted and modified (see the original ISO 15926) would we
see the end to the standardization processes?
MW: Not for a long time. I expect that for Oil and
Gas we will be adding to ISO 15926 for at least 10 years. There are incremental
benefits to be gained as more is added (i.e. you do not have to wait ten
years). The question for me is "When do you have enough critical mass to
present a compelling business case?" I can only say "not yet".
I
mean, the end to uncertainly. I do not
mean a closed system where every element is set forever.
MW: Not sure what this means. You always rely on
people using the ontology correctly. As one of my colleagues says "You
can't stop dumb people doing dumb things" which he puts together with
"we are all dumb at leas some of the time".
MW: But I'm an engineer, so my aim is better, not
perfect.
I
would also like to point to OASIS standards
http://www.oasis-open.org/specs/index.php
in
particular to
http://www.oasis-open.org/committees/ebxml-cppa/documents/ebcpp-2.0.pdf
which
is an detailed set of specifications for electronic business XML collaboration
protocol profiles.
for
example:
The
role element has the following attributes:
a required name attribute
a fixed xlink:type attribute
a required xlink:href attribute
I
ask again, why the standardization process is not coming to a completion, given
that specifications like that OASIS specification were largely completed in
2002.
see
also
http://www.ontologystream.com/beads/nationalDebate/201.htm
I
am reminded of John Sowa's recent note to me regarding the need to legislate
standards.
In
a note from
http://www.ontologystream.com/beads/nationalDebate/206.htm
we
do not see any reference to OASIS. Why
is this?
In
summary, a reference ontology seems to be about concepts used within a domain
space, (I am thinking about the domain of all e-commerce activity). So the
http://www.oasis-open.org/specs/index.php#cppav2
could
be a reference ontology that has specific specifications that allow
interoperability between "things" that need to interoperate. Why is this not the end to the
standardization process?
Again,
the imposition of a "logic" or something like lattices for use in
inferencing, seems to be the imposition that kills clarity.
A hub of reference models is what the B-2-B
transaction space needs. Yes?
MW: I think so, with mappings between them.
But I don't think this will bring the end of standardization. The nice thing about standards, as you effectively point out here, is that there are so many to choose from. The reason for this is that invention is the mother of necessity. It is often easier to reinvent something than to find the original invention and use it. People also just have different ideas about how to do things, so they go off into their own part of the woods and do it their way. This is a good thing in something as immature as ontology development.
As the area matures, there will be a natural aggregation around one or two of these developments. I think the best we can do is to provide an environment in which this can happen. What I do not think we should do is to pick one and stifle development of others (one of which might have the seeds of where we really should be going).
MW: I think an
ontology hub would encourage aggregation, whilst enabling differentiation. It
will encourage "stealing from the best with pride" and make
reinvention (especially when worse than the original) unnecessary and expensive
compared to reuse.