Re: Thoughts on automatic discovery

Massimo Paolucci wrote:

> A paper on discovery using DAML-S is available at
> http://www.cs.cmu.edu/~softagents/papers/ISWC2002.pdf

Massimo,

This interesting work may have a flaw. (Or is it just a blemish?)

One of your arguments (in section 4) is that semantic matching is better
than UDDI's matching because "UDDI does not provide any support for
finding services on the basis of what tasks they perform."

An example of such a task is 'car selling', which is the task that a
'car selling service' performs. An ontology that included this task
could have 'car leasing service' as a related task, and something like
'car providing service' as a broader task.

The problem is that the example you use in your paper doesn't appear to
use any ontology of tasks to do its work, instead it uses an ontology of
things: 'vehicle', 'car', 'SUV', etc.

So my question is this: do you actually use an (unmentioned) ontology of
tasks in your system? And if not, then doesn't this weaken your argument
against UDDI?

Thanks,

Jeff

Received on Tuesday, 23 April 2002 13:55:09 UTC