[Pellet-users] ANN: Owlgres 0.1 (Alpha Release)

[Bijan Parsia]

On Thursday, May 8, 2008 2:33pm, Kendall Grant Clark <kendall at clarkparsia.com> said:

> Excerpts from James Howison's message of Thu May 08 13:58:02 -0400 2008:
>> This looks interesting.  Assuming your ontology fits its capabilities,
>> the advantage of this over full Pellet, and for example the Jena
>> reasoners, would be that because it is using the database there is no
>> need to load the whole ontology into memory (because it is using
>> postgres?)
> 
> Yes, the advantage is being able to reason over KBs that exceed the size of main
> memory. That is, scalable abox reasoning for DL-Lite KBs.

That and reasoning is generally (or should be) faster and more robustly scalable. Faster in that the worst case complexity (esp. data complexity) is much lower and the implementation is far simpler. More robustly scalable in that, in principle, you should be able to through arbitrarily complex DL Lite ontologies (and data) at the system and get reasonable behavior (whereas, due the nature of OWL, it's not that hard to build something that requires new optimization or tuning).

Robustness (for the effort) has been reasonably seen in EL++ reasoners, and I feel pretty confident that you'll see it with DL Lite. If you have a really deep class heirarchy of a certain sort and just the right queries, it's possible to get into trouble (because of the query expansion process).

It'll be interesting to see how it works out in practice. We *really* need good test scenarios! If anyone has potential applications that would fit and needs help getting it together, please let us know!

Cheers,
Bijan.