In my experience, we and
our machines use ontologies to describe and interact with the physical world.
This means that for ontologies to be useful they have to have some accompanying
physical semantics, not just a translation into some other concepts or words, or
even an axioms to describe their meaning. I am of course strongly in favor of
axiomatic semantics for ontologies. But even with axioms there need to be
procedures for constructing or recognizing if something is in a specific
category. One might call this an operational semantics. This is model theory in
the sense of logic, but it does not come for free, even when there is an
axiomatic semantics.
Taking “if it walks like
a duck, …” a bit further, a vehicle design description may use an ontology with
categories such as physical object, with specializations for some particular
kind of steel. The physical semantics is generally a procedure to recognize or
test for that kind of steel. A big
problem for replacement parts for commercial aircraft is the presence of
counterfeit parts. Often these recognition procedures, as well as procedures for
performing tests and analyzing test results, are quite complex, with their own
ontologies to describe them. However, these ontology applications are only
successful when the physical semantics can be agreed on in the community where
the ontology is used.
Practically the
successful use of an ontology depends on having a well-defined explicit
operational semantics. It is the operational semantic issues that cause the
tears, and litigation. This does not necessarily mean that everyone agrees with
the definition, but they understand how the concepts are being used in a given
context.
I presume that the
customer who takes possession of an oil refinery checks to see if it has all of
the equipment which the customer is expecting. I know this is the case with
other large expensive industrial products. If some regulatory agency is faced
with closing down a bank, or a food processing plant, the issue is likely to be
how do we measure the conditions which trigger action.
I am not sure that this
aspect of ontology use has gotten much attention in the recent
discussions
Henson
Graves
Sent: Wednesday, February 19, 2014 7:23 PM
Subject: Re: [ontolog-forum] Good ontologies without good tools are
useless
Hi Henson and all,
We are finding similar
requirements in finance, as I'm sure Mike Bennett would echo. The
ontologies that are needed to support compliance and understanding the nature,
risk and opportunities related to complex securities, annuities, and other
financial instruments are necessarily far richer and more rigorous in their
definition than what I have seen over time from a linked data / big data
perspective (although the systems that do or will use them are quite large and
the data volume certainly qualifies as big data). The services that are
needed to support both regulators and bankers require reasoning of various
sorts, including but not limited to rule-based inference, classification, and
machine learning. Systems to support fraud detection, anti-money
laundering, insider trading detection, and many other processes have been using
rule-based reasoning for decades.
Use of UML for software engineering has
been fairly widespread in banking and insurance for many years, and retraining
information architects in those industries to use it for ontology development
(with ODM stereotypes) is not as big a leap as attempting to get them to use
Protege or other ontology tools. Another advantage is that the UML models,
which can be exported to OWL or some other rule language, are available
immediately for reuse in the software applications folks are developing.
This minimizes reinvention of the vocabulary used in development. Use of
the same ontologies in rule-based systems, again rather than reinventing the
vocabulary, also dramatically reduces error, especially if the methodology
includes analyzing the ontologies to ensure they are logically consistent prior
to reuse.
For those that are interested, we are in the process of
publishing a new version of the ODM at the OMG, which addresses a number of
deficiencies we've found in implementation over the last several years.
ODM 1.1 supports most of RDF 1.1 and provides better support for OWL 2, although
the work there is incomplete. We plan to have a 1.2 version of the
specification by sometime this summer that will fix the remaining known problems
and provide complete coverage of OWL 2. Both specifications and related
artifacts will be available from the OMG site once they are approved by the
membership. That process is largely complete for ODM 1.1, with a couple
more process steps remaining, but the spec should be public soon. If I had
to guess, I would say the 1.2 version will be available by this fall, if not
sooner. We considered waiting to publish until we were "done", but so many
changes had already been made that we really needed a new baseline to work
from.
For OMG members, the Ontology Development Metamodel (ODM) 1.1
Revision Task Force (RTF) report, documents, and all of the model artifacts are
already available on the ODM 1.1 RTF work in progress page. They should
become public when the process hoops are complete, possibly by sometime next
month.
Members of the ODM revision task force include some of the same
folks who have been working on the SysML specification, and who really
understand the value proposition, as Henson puts so well below. The task
force also includes a couple of the major UML vendors -- No Magic and Sparx --
and there are beta tools for both No Magic's MagicDraw and Sparx EA
available.
Best regards,
Elisa
On 2/18/2014 8:52 AM, henson
wrote:
I work with engineers who have a day job and recognize that they need
ontologies to do their work. [I have a lot of experience with being in their
shoes.] The ontologies are needed to model, i.e., describe and specify systems
and their operations in the real world. The models are used to design and
analyze vehicles, aircraft, etc. The recognition of the need for ontology is
that the operating environment descriptions need to be much more complex and
are much more changeable than they were 50 years ago. The use of ontology also
increasingly applies to some of the systems being built. They use ontologies
to process the enormous volume of data that they ingest at operation time, and
use some inference to take actions. The software of some of these
systems contains a model of the system as well as its environment which it
uses at runtime for flight control and threat avoidance.
To be of any use the ontologies have to be imported into the development
tools which they use. The mostly UML based tools (including SysML) are
now robust, ordinary engineers can and do use them to develop large complex
systems. By and large these folks cannot or will not use traditional logic
syntax. Also there are not commercial grade tools that have been proven in the
industrial context. This doesn’t mean that these language don’t need a formal
semantics, and need extensions to handle the applications that they are being
used for. They do. I have been arguing for a long time that the formal methods
folks should focus on retrofitting and evolving these tools, rather than
attempting to develop new ones.
As a practical note, I believe that the SysML community is more receptive
to something such as William Frank advocates than the UML community. The
reason being that sociologically the engineering community has to deal with a
much broader scope of applications than the UML community. If the battery fire
on a commercial aircraft causes a crash then the manufacturer will certainly
be sued and will be asked to produce the analysis and test results that were
used to declare the aircraft safe to fly. If these results are not compelling
the manufacture is in big trouble. Engineers are beginning to get this.
As a comment on Ron Wheeler’s comment the ontologies I see in the big
data world if they deserve the name ontologies, are currently much simpler
than the kind of ontologies mentioned here. However, if these systems are to
be used for medical diagnosis, and drug design and analysis then they will
have to have the complexity of the ones I am talking about.
I do not know of anywhere within a university context material relevant
to this discussion is being taught. There presently doesn’t even seem to be
any traditional departments willing to pick this up, in my limited
experience.
By the way, I am passing along John’s slides to a group I am working with
which is in desperate need of an upper ontology.
Henson
Sent: Tuesday, February 18, 2014 9:15 AM
Subject: Re: [ontolog-forum] Good ontologies without good tools
are useless
Yes,
UML based on a many-sorted higher order predicate calculus
with Henkin semantics, and a simple upper ontology represented by the
sorts -
this is what we (Joauquin Miller, Kevin Tyson, and
I) proposed for UML 2, in Clear, Clean, Concise (3C) UML
Communications of the ACM, Nov 2002, volume 45, no. 11 pages 79 -
81.
"The Clear, Clean, Concise (3C) UML2 proposal makes the
language
easier to understand and enables it to describe a broader
range
of systems, from Web agents and services to entire
business
communities."
This was never going to happen, at that
point. The agenda was set by Oracle and IBM, with no regard for the
benefits to the long term future of systems engineering. Also, I
misunderstood people so much that instead of referencing the science, I simply
explained the BENEFITS, and showed how simple it was to use this
language, so I now suspect they thought this was some new off-the-wall
approach invented by us three. I agree with all you
say below, John, but add that an equally important feature integrated into a
cosistent set of of UML models, along with the 4 you list, are state -
transition models, the backbone, in my opinion, for precise behavior
specifications. Myself, I find UML diagrams, with my OWN simple
simple common logic semantics, the most effective way to ensure a system
consistency, because of all these integrated models.
_________________________________________________________________ Message
Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxxShared
Files: http://ontolog.cim3.net/file/Community
Wiki: http://ontolog.cim3.net/wiki/ To
join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
_________________________________________________________________ Message
Archives: http://ontolog.cim3.net/forum/ontolog-forum/ Config Subscr:
http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/ Unsubscribe:
mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx Shared Files:
http://ontolog.cim3.net/file/ Community Wiki: http://ontolog.cim3.net/wiki/
To join:
http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
|
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (01)
|