Ontology-based queries

[Sean Bechhofer]

On Tue, 15 Oct 2002, Ashish Mahabal wrote:

>
> Tony wrote:
>
> >On the other hand, if we modify the above query to be:
>
> > Object is-of-type SNR
> >  AND Source registers-in-wavelength X-Ray
> >  AND Object is-within (1 arcmin) of-region (05 25 00 -69 38)
>
> >I think we hit difficulties. I cannot see how an ontology-based query
> >like this could work.

OWL will include support for (unary) datatypes, so this kind of
expression/query can be represented.

> The above query is likely to be expanded out to:
>
>  Object is-of-type SNR
>   AND Source registers-in-wavelength X-Ray
>   AND Object has-RA > x.y
>   AND Object has-RA < z.w
>   AND Object has-DEC > a.b
>   AND Object has-DEC < c.d
>
> But then, as you imply, the question of a subclass versus individual
> objects would arise and the query may have to be split into two sections:
> apples and oranges. Do just the first part first using an ontology:

There's no reason why an OWL query engine would not be able to answer
queries like this. As Ashish points out, the query is likely to be
answered through a division of labour between the ontological part and the
datatype part. OWL makes a clear separation between datatype and object
properties, and there are proposed architectures for reasoners (see [1])
that will support precisely this kind of query.

>  Object is-of-type SNR
>   AND Source registers-in-wavelength X-Ray
>
> and then on the resulting set do the SQL like query
>
> select * where Object has-RA > x.y
>   AND Object has-RA < z.w
>   AND Object has-DEC > a.b
>   AND Object has-DEC < c.d
>
> However, if we can store the classes (SNR, X-ray) with an object along
> with its attributes (RA, Dec), we could do the two in one fell swoop.
>
> Whatever way is chosen, an important point to remember will be that the
> query should magically rearrange itself so that the first clause is likely
> to return the smallest number. Otherwise if we go through a large number
> of objects only to throw away most of them at a later point, its a waste
> of resources. But perhaps this point will come up much later.

Of course, query optimisation is essential. I would hope that this would
be part of the reasoner's responsibility, so a client application would
not have to worry about this (much as is the case with standard query
engines).

	Sean

[1] http://www.cs.man.ac.uk/~panz/Zhilin/download/Paper/Pan-Horrocks-datatype-2002.pdf
-- 
Sean Bechhofer
seanb at cs.man.ac.uk
http://www.cs.man.ac.uk/~seanb