Lists Home |
Date Index |
Jonathan Borden wrote,
> Miles Sabin wrote:
> > I don't believe that any logic that's likely to be interesting in
> > this kind of context is decidable. Let's face it, your semantic web
> > reasoner is going to have to do the same job as the transform, so
> > how could it be guaranteed to terminate in finite time if the
> > transform can't be?
> Ahh but that is _exactly_ the point. DLs are carefully crafted
> exactly so as to be able to give you this guarantee. DL is _not_
> Turning complete. The problem of classification does not require a
> Turing complete processor.
Ahh, but it's exactly _my_ point too ;-)
The fact that DLs are so carefully crafted means that their scope is
pretty drastically limited. I think that it's quite likely that this
lack of expressiveness will cause major headaches in many, many
contexts: things which can be expressed quite deftly with arithmetic or
a sprinkling of second-order machinery, can be either inexpressible or
explosively long-winded in weaker systems.
How would you say "X has the same properties as Y, apart from P" or "X
has twice the number of widgets as Y" in a DL? Aren't those the kinds
of things you'd want to be able to say easily when expressing a mapping
Nb. I'm _not_ saying that DLs aren't useful ... only that the
translation from an intuitive understanding (which is very likely to be
riddled with arithmetic and second-orderisms) of a mapping problem to a
functionally equivalent DL rendering is often going to be distinctly
non-trivial. In those cases you might well want to trade off
decidability for expressive power.