|
Bill,
You are probably right - we could
well have met at a PDES meeting. But in my defence, there are nearly half a
million hits for William Burkett on Google. There was no easy way of checking
that the .bah.com domain had the same semantic reference as the old
.EuroSTEP.com domain, even had I thought to check.
I'll agree that standards for
engineering are more precisely defined than standards for, say, beauty, and
therefore we cannot expect the precision of semantic matching to be the same. I
will also agree that testing for conformance can never be exhaustive. However,
ultimately what this forum is about is the engineering of computer systems,
rather than the richness of poetry. The question is how do we apply traditional
quality standards that we use in written communications to computer
interpreted data?
If I write a letter offering for
sale a bolt conforming to a particular design, the purchaser will check that I
am a reputable supplier, and that therefore the bolt should match the
specification. There is also an objective measure whether the bolts I supply
meet that specification, and legal recourse if they don't - and the standard
will be used as part of the evidence. The question is, how we contextualise
ontology assertions so that we know whether we can rely on anyone else's
assertion in the same was as we currently do for paper (here I consider most web
pages as electronic paper, as the function of the computer is to display rather
than interpret).
Oddly enough, yesterday I took a
phone call from someone I had not previously met, but was interested in work I
am doing on ontologies for metadata, and the conversation turned exactly to this
topic. I'm not saying that this is the hot topic of the moment, but it seems to
be coming up up people's agendas.
Sean Barker
P.S. (I just checked LinkedIn and
there are 69 people with a matching name - although you come out
number 1 in the list).
*** WARNING ***
This message has originated outside your organisation,
either from an external partner or the Global Internet.
Keep this in mind if you answer this message.
>?
although
we have never corresponded before ?
I think
we have corresponded before, Sean ? I think we actually even spoke to one
another when I was with Eurostep. Do I misremember? J
As for
your points, I?m not disputing the value of standards. I?m disputing that
is possible to certify the semantics (i.e., meaning) of the use of a particular
data element, i.e., that a data value and use of the data element mean what the
definition says it?s supposed to mean. Standards for mechanical things
(e.g., ?volt? and ?bolt?) are very different than standards that involve
language, interpretation, and judgment.
I think
your use of the term ?semantic standard? differs somewhat than my interpretation
of it. At one point in my life I would have called STEP/PLCS and other
such standards ?semantic standards?, but no longer think those words belong next
to one another. Standards like STEP/PLCS are data exchange standards, and
you can absolutely test syntactic conformance to the standard. You cannot
test semantic conformance to the standard, however; at best, you can
verify through mock usage (i.e., testing) that two parties are interpreting the
standard compatibly.
Which
you?ve essentially said in your response: ??tests conformance ?to make sure that
new versions of software can interoperate using the standard, and to test that
the vendors have the same interpretation of extensions to the
standard?. That?s catching and correcting semantic errors by testing
for articulation and interpretation errors before they?re uncovered in
use. I don?t think this testing will ever be complete enough, though; as
the standard ages and business requirements change, new and more semantic errors
will occur and the standard will have to adapt.
Perhaps
my original objection was based on a misinterpretation of the words you chose to
express your original point. J
>The
fact that we can have this conversation ? is a testament to the effectiveness of
semantic standards,
But even
given the richness of natural language, it?s still very easy to
mis-communicate. I doubt it will be any different using data to
communicate.
Bill
From: sean barker
[mailto:sean.barker@xxxxxxxxxxxxx]
Sent: Monday, October 26, 2009
1:55 PM
To: ontolog-forum@xxxxxxxxxxxxxxxx
Cc: Burkett,
William [USA]
Subject: Re: [ontolog-forum] Wish list for the sematic
web
Bill
The
fact that we can have this conversation, although we have never corresponded
before, is a testament to the effectiveness of semantic standards, starting
with the definition of the Volt, the way you wire up the plug on our
computers, and all the other thousand standards that contribute to the
mechanics of communication. When you buy something on line, you rely on having
common semantics for the transaction, starting with the meaning of "$" though
to the supplier telling the postal service your zip code (and them
understanding it in terms of the geography of the USA).
The
answer to your objection comes in two parts:
Firstly,
there are already some established ways of checking conformance to semantic
standards. The CAx-IF regularly tests conformance of mechanical CAD systems
for their conformance to STEP (ISO 10303), both to make sure that new versions
of software can interoperate using the standard, and to test that the
vendors have the same interpretation of extensions to the standard. I
would imagine that similar testing goes on in the commercial sector, for
example, for EDIFACT messages. The objective is not to prove conformance to
the standard ab initio, but to confirm conformance to the standard, probably
after several rounds of bug-fixing.
Secondly,
whether or not you require proof of conformance to a semantic standard is a
measure of how much risk you are prepared to take. You may not require proof
that the Hawaiian Pizza you are ordering actually has ham and pineapple, but
if it doesn't it will have only cost you $10 and some space in your garbage
can. A bolt on an aircraft might also cost only $10, but it might also cost
the lives of hundreds of people. The aerospace industry is rightly very choosy
about who it trusts to make bolts, and the manufacturers have to go through a
relatively expensive certification process.
Just
to put this in context, as a major world supplier of defence systems, we
supply hundreds of products with millions of components to tens of customers,
each of whom also has hundreds of suppliers, as indeed do we. The PLCS
standard is being put in place to simplify the support by using a common set
of semantics. The problem is not to implement thousands of slightly different
PLCS sysstems, but have common systems with common semantics. (BTW the
standard is written as a conventional Entity-Relation schema plus a
taxonomy to specialise generic elements and rules to tie down particular
transactions - almost RDF).
The
problem is not that as human beings, we cannot negotiate the meaning of our
communications, but rather, whether we can trust our computers to do some of
the work for us, without manually checking each step that they have taken.
There was a report a year or so ago that someone had developed autonomous
agents that had then gone on to develop their own language for communicating
with each other. The problem was, nobody knew what they were saying. I do not
believe this is the aim of the semantic web.
Sean Barker
BAE SYSTEMS - Advanced
Technology Centre Bristol, UK
+44(0) 117 302 8184
BAE Systems
(Operations) Limited
Registered Office: Warwick House, PO Box 87,
Farnborough Aerospace Centre, Farnborough, Hants, GU14 6YU, UK Registered in
England & Wales No: 1996687
From:
ontolog-forum-bounces@xxxxxxxxxxxxxxxx
[mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Burkett,
William [USA]
Sent: 26 October 2009 12:59
To:
[ontolog-forum]
Subject: Re: [ontolog-forum] Wish list for the
sematic web *** WARNING *** This message has originated outside your organisation, either from an external partner or the Global Internet. Keep this in mind if you answer this message.
Sean:
In
my personal opinion, based on my personal experience, your list starts off
with a major misstep:
>1)
An examination of methods for certifying that the user semantics for terms
matches the definition of the term.
I
don?t believe there is any possible way to do this. The semantics of
data and the schemas that govern them are created by people (or their software
surrogates) using their native linguistic skills, - complete with all the
language foibles and inconsistencies that plague human-to-human
communication. Consider how much time and money is spent creating legal
documents and arguing court cases over what the language in a legal document
means ? I think the same interpretative imprecision and vagueness will be
present in most data. Note that dictionaries do not assert what a
word means (nor do they enforce that meaning), but rather dictionaries
document the conventional usage of the term.
I
think the main objective of semantic technologies needs to be semantic
adaptability: the ability to respond quickly to and correct semantic errors ?
much the same way we learn the idiosyncrasies of the way each of us speak and
adapt to them.
Bill
|