[seek-kr-sms] FW:update- from ferdinando on workflow language

Serguei Krivov Serguei.Krivov at uvm.edu
Wed Nov 3 11:42:40 PST 2004


Hi Serguei - thanks for the updates (and for spreading the Word). I'm
busy with meetings with Conservation International so I have little time
for feedback. Anyway I don't think that OWL would necessarily be a good
direct substitute for MOML - a suboptimal one perhaps, because capturing
control structures in axioms isn't necessarily the most natural thing,
but on the other hand MOML is so primitive than the simple link to
ontologies would be enough of a reason to switch, as you pointed out.
As you know I do think that 1) a KR formalism is best to capture the
conceptual essence of what you want the workflow to represent and
produce; 2) with the proper upper ontologies it's possible to map that
concept in a workflow; and obviously that (3) the concept is the closest
match to what the user thinks when s/he develops the
model/workflow/pipeline. Particularly if it's represented with an
intelligent, domain-optimized graphic that could be a GrOWL panel
extended with your graphical extensions. Think of stella - it's popular
because the boxes "are" populations or carbon stocks, not because a
stock plus a flow is a difference equation. Kepler will show you the
differential equation, although you can make somersaults to make it look
like stella. In OWL, you can model a lotka-volterra predator/prey
community as such - then proceed to specify that the individual
population's state can be represented as an equation that depends on the
others, and take it from there. All you change from the DATA
representation of a community is the definition of the state as an
equation rather than a number - change over time - and you can use an
existing measured community (EML-tagged) as the data template for a
dynamic model, or store the data and the model in the same object -
running the model becomes just "projecting" a dataset over time.
(That's roughly the demo I want to make in the next few months).

What I think would be a great strategy is to add a concept panel to
Kepler, and have a possibly improved MOML workflow generated by a
semantic "director" (which I have partly implemented in IMA), while the
generated workflow remains accessible to the user as before. I think
both you and I would be happy to work on something like this with the
other Seekers.

Anyway - you know all this. Feel free to circulate this message if the
circumstances are appropriate. Thanks for the updates. Ciao, ferdinando

 




More information about the Seek-kr-sms mailing list