Re: Can Shapes always be Classes?

It appears to me that SPIN can handle this kind of conditional constraint 
quite well.  All that SPIN needs is *some*  type that can be used to kick off 
the constraint evaluation.   After that, other information can be used to 
determine whether the constraint is satisified, for example, type Issue could 
have a constraint that says  (roughly) if there if the value for the status 
property is closed then there has to be a value for the resolution property.

At the extremum, one could just attach SPIN constraints to rdfs:Resource and 
have the "body" of the constraint do all the conditional processing required 
so no explicit type arcs are required at all.  (Well maybe not in SPIN as 
implemented but we can imagine a SPIN that is RDFS-compliant.) This 
methodology might need some fancy optimization to work well in practice, however.


peter


On 11/06/2014 09:59 AM, Eric Prud'hommeaux wrote:
> * Irene Polikoff <irene@topquadrant.com> [2014-11-06 10:46-0500]
>> True - if you have data without any type triples, they would need to be inferred first. Not a problem and, furthermore, in my experience, this situation is atypical.
>
> _:Bob has a type arc (foaf:Person), but that type doesn't fully
> discriminate all of the possible constraints to which it may conform,
> as would be the case with pretty much any publicly-consumable data.
> This is less an issue for proprietary data confined to a known number
> of business processes.
>

Received on Thursday, 6 November 2014 18:20:10 UTC