Lists Home |
Date Index |
Note that I'm talking about the "semantic web" of actual technologies
not the Semantic Web of certain visionaries ... and it's more respect
than actual love.
On Jun 12, 2004, at 12:32 PM, Jonathan Borden wrote:
> SNOMED has been standardized as a medical ontology which as been
> licensed by both the U.S. and U.K. governments for use in healthcare.
> I'd say that this stuff is being used. Dismiss this at your own
> discretion. Perhaps it doesn't affect your corner of the world.
Jonathan mentioning SNOMED a year or two ago in a previous incarnation
of the permathread was the "aha!" moment for me. There are certainly
corners of the world where systems of terminology/ontology are well
enough established to make even the kinds of inferences possible with
OWL/DL useful. I believe the example he used back then was something
like querying a database that had no specific code for 'brain tumor'
for all cases that involved a type of tumor occurring in the brain. If
each case does have a SNOMED code associated with the diagnosis, a
reasoner can infer whether or not that is a kind of brain tumor,
without human intervention or an exceedingly complex query.
Reasonable people can disagree over how many cases there are where a
well-defined ontology exists; after all, the SNOMED folks just
formalized the ontology implicit in medical theory and exploited
nomenclature widely used in practice. That took a couple hundred
(thousand?) years to get to its current state, and human physiology and
(to a lesser extent) pathology hasn't evolved all that much during that
time, so knowledge wasn't rendered obsolete every generation or so.
My second epiphany about this stuff came more recently -- it became
brutally clear that internet, XML and web services technologies had
done a lot to remove the mechanical barriers to data interchange, so
exchanging well-understood document, data records, and service
invocations across platforms is no longer the painfully labor
intensive proposition it was even a decade ago. Now that the plumbing
is in place, however, it is clear that the barriers to effective
communication lie more in what the data *means* than in what format it
is in or what protocol will be used to exchange it. One might hope
that industry-wide working groups will sort out the differences for
each vertical.Wwheeooooffff [sound of dope smoke being inhaled ;-) ]
One might hope that people will value interoperability more than
inertia and adopt something like UBL [Kumbaya .... Kumbaya]. One might
anticipate that some Omnipotent Entity such as the US government,
WalMart, or Microsoft will just enforce uniformity [could happen, but
the proles tend to resist such attempts by Big Brother].
One might much more plausibly believe, IMHO, that a) individual
organizations can formalize what *they* mean by various terms,
namespaces, etc. by reference to concrete documentation that describes
them or software components/database fields that implement them; and b)
that these private ontologies could be shared and mapped-between by
those needing to exchange data across organizational boundaries. Maybe
someday those will evolve into shared ontologies such as SNOMED, we
shall see, but we don't need to believe in such things to use OWL, etc.
to formalize and manipulate the private taxonomies/ontologies that are
in actual use.
The objection from the "XML is all you need" contingent seems quite
well taken to me: At the end of the road, it comes down to bits on the
wire that are being matched and manipulated. It may be that one can
effectively cut out the ontological abstractions and deal directly with
the syntax patterns and transformations (as XQuery and XSLT support
quite well) in your domain of choice. My guess is that there is enough
useful higher-level structure in natural language and the real world to
give the exercise of building taxonomies/ontologies some real value in
a lot of application domains now that there is a bit of a network
effect around the "semantic web" [lower case!] tools that will make
these things useable by ordinary programmers and human end users.
The OWL Guide begins: "The World Wide Web as it is currently
constituted resembles a poorly mapped geography. Our insight into the
documents and capabilities available are based on keyword searches,
abetted by clever use of document connectivity and usage patterns. The
sheer mass of this data is unmanageable without powerful tool support.
In order to map this terrain more precisely, computational agents
require machine-readable descriptions of the content and capabilities
of Web accessible resources" I guess I've stopped worrying about the
grandiosity of this vision (and the fact that we're muddling through
with all sorts of little things like Google and WSDL fairly nicely),
and learned to respect the value of machine readable descriptions of
local resources, and the potential this has for reducing the complexity
exposed to developers and users. We shall see ...