Re: OWL tools Re: Tools for 20 million triples?

Dan Brickley wrote:

>* Charles McCathieNevile <charles@w3.org> [2004-03-27 22:20-0500]
>  
>
>>Hi folks,
>>
>>So it seems there are plenty of systems capable of handling lots of RDF. Are
>>there any that handle OWL (in some version or other) at that scale?
>>    
>>
>
>I guess it depends what you mean by 'handle'.
>
>OWL descriptions of RDF vocabularies (such as FOAF) license inferences,
>ie. justify you (or your software) concluding new things from a 
>dataset. Tools like Foafbot (http://usefulinc.com/foaf/foafbot) which
>are built on top of generic RDF machinery (in this case Redland) can do
>useful things (justified by OWL) without being complete OWL reasoners.
>For eg. FOAFBot's Web crawler does a bit of 'identity reasoning' based
>on knowing that some but not all FOAF/RDF properties are 
>owl:InverseFunctionalProperty, ie. uniquely identifying. 
>
>So the good news is that OWL support can be added incrementally. You
>might decide just to support an understanding of inverse properties, 
>eg. making a system understand that foaf:depicts and foaf:depiction say that
>same thing as each other, just expressed inversely.
>
>  
>
Which leads in an interesting direction. If IFPs are particularly 
important in your system, you may want to optimise around them without 
worrying about efficiencies elsewhere. There's a Horrocks paper [1] on 
optimising DL subsumption. Perhaps we could do with exploring/compiling 
a list of some comparable (best practices?) possibilities on other 
optimisations - identity resolution, disjoint sets etc.

>The bad news is that I don't have any handy references to systems that
>do this sort of thing...
>
>  
>
Nope, me neither.

Cheers,
Danny.

[1] http://www.dbnet.ece.ntua.gr/~dwq/p19.pdf

-- 
----
Raw
http://dannyayers.com

Received on Sunday, 28 March 2004 12:01:19 UTC