[Pellet-users] A lot of descendant queries

[bijan at clarkparsia.com]

On Thu, May 3, 2007 2:25 pm, Alan Ruttenberg <alanruttenberg at gmail.com> said:

>  On May 3, 2007, at 2:21 PM, bijan at clarkparsia.com wrote:
> 
>> That being said, you of course could reflect things into a form
>> where this could make sense. But I'm unclear on what the gain is.
>> Maybe I don't understand your point?
> 
> The point was that if it performed the calculation I needed,

If *what* would do this? I think I'm missing some pieces. Do you intend to create a new property "is-a", populate an abox with is-as reflecting the fully classified class graph, etc. etc.?

> and it
> were faster than the current methods I was using,

I'd be surprised if anything current would be faster than something akin to the classification style code Evren suggested. Whether it's faster than what you are currently doing is unclear to me too :)

> then it would be
> useful in solving my current problem: I need the relationships. Also
> that it might point to an optimization for the sparql dl compiler -
> recognizing the form of the query matched a simpler problem and using
> existing code to implement the solution.

Well, this is, indeed, a general optimization strategy, and I suppose it's imaginable to encode the problem in such a way that the automata code did a very good job...I guess. I doubt it'd be so straightforward. But I'm a little sleepy so what do I know :)

Cheers,
Bijan.