Mills, Mike, and Ron,
Let's assume SME's, business users, and IT can build something > sustainably that's ontology driven, will anyone come?
I believe the major task is not to build something from scratch,
but to build something new that co-exists with existing systems.
There are trillions of dollars of legacy software that runs the
world economy, and it's not going away any time soon.
Ideally, any new ontology should be based on the way the business
works -- that means a large part of it should be *derived from*
The diagram looks a little scary to say the least.
Have people moved some of the semantics "heavy lifting"
into the area of queries and does this point up some
disconnect in the use of ontology itself?
I wouldn't say that it's a problem in the ontology itself,
but in the way that ontologies are developed and connected
to the applications.
That is an issue that we addressed at VivoMind. We focused
on automated and semi-automated methods that derive the KB
from whatever data we are given -- structured and unstructured.
For some examples, please look at the applications discussed
in the following slides:
You can jump to slide 14 for the list of 3 applications.
The first one, Educational Software, shows how we used the
VivoMind Analogy Engine to allow the system to adapt to
the users (slides 15 to 24).
The second application, Legacy Re-engineering in slides 25 to 33,
shows how the system derived the data model and the knowledge base
directly from (a) the COBOL programs and (b) the documentation
about those programs.
For this application, the customer just wanted a CD-ROM that
contained the results of the analysis. But the same approach
could be used to connect any new ontology to the KB that was
automatically derived from the legacy software.
The third application, Oil and Gas Exploration in slides 34
to 41, shows how the system can learn new knowledge by reading
a textbook. See the concluding slide 41. Then go back to
slides 38 to 40, which show how knowledge derived from the
textbook is used to form new links between the user's query
and research reports.
... the promise and premise of IBM's technology used to construct
Watson ... is supposed to integrate the "definiteness" of ontology
with the insight drawn from conflicting and ambiguous knowledge
floating loose in academic papers and popular media.
That's what VivoMind is doing for the examples in the pursue.pdf file.
In the discussions about Watson in Ontolog Forum, nobody was able
to discern anything remotely similar in any of the publications
from IBM. If they had it, I'm sure they would have emphasized it.