Re: [TF-ENT] URIs for entailment regimes in service descriptions

2009/11/1 Sandro Hawke <sandro@w3.org>:
>
>> one of the problems with just saying OWL is that reasoners that use
>> Direct semantics only work on a subset of RDF documents (those that
>> fall into the OWL DL syntactic fragment). Most will not even parse
>> non-OWL DL documents and OWL EL reasoners might not even parse OWL DL
>> documents.
>
> This is where "repair" comes in.  It's not standardized, but it doesn't
> have to be.  My hope is that OWL DL reasoners will modify the graph to
> something they can reason with, thereby becoming sound (but incomplete,
> of course) RDF-Based-Semantics reasoners.  (They also have to do a few
> tricks detailed in [1], I guess.)  If they do this, we'll have a whole
> lot more interoperability.

Hm, we have no such plans for HermiT. At least I don't and we have
never talked about this as something on our to-do list for HermiT.
Pellet does at least a little bit in that direction if I recall
correctly.

> In particular, for the market, I believe it is crucial that users who
> are not expert ontologists never have to pay attention to which profile
> or semantics are being used for OWL.  I don't yet know the best way to
> motivate implementors, though.

Neither do I.

>> Since these URLs are for the service descriptions, it is not the user
>> that specifies this.
>
> Hmmm.  Isn't there also a proposal to allow SPARQL clients to ask/allow
> the server to do reasoning, some way for clients to turn on/off the
> reasoning in the server?  [ Sorry for not being up on the drafts yet ]

So far, the endpoint declares in its service description what
entailment regime that endpoint will use. This is per endpoint and
each graph within that endpoint is used with the declared entailment
regime. Andy in particular is in favour of a finer grained
description, i.e., for each (named) graph in the data set, the
endpoint declares which entailment regime is applied to that graph.
The user gets, in a sense, what the endpoint declares to use. If I
query endpoint A and A does RDFS entailment, then that is what I get.

> For service description, what profiles and semantics should be used to
> describe a reasoner which does DL but can also handle some Full?  All
> profiles, and both semantics?  To correctly handle the little cases
> where the semantics differ even within the DL subset, the client's going
> to have to say which one to use, I think.  (Or the service descriptions
> will have to name different graphs within that endpoint, or some such
> trick.)

That can, at the moment, not be described.
At the moment, a client can choose to use and endpoint or not. If I
wanted to offer HermiT as an incomplete OWL Full reasoner, then I
could offer this service under a different endpoint URI and describe
that as OWL Full endpoint. There is no way though to declare that it
is incomplete.

There is a whole lot of things that one could do or woud want to do
for entailment regimes, but since it is a time permitting only feature
the premise was so far to impact other parts of the spec as little as
possible. My goal is to get entailment regimes defined properly. If we
think that adding more advanced descriptions for entailment regimes to
the service descriptions document is essential to make them usable,
then ok, but otherwise I tend to say let's keep it as simple as
possible.

Birte



>   -- Sandro
>
> [1] http://www.w3.org/TR/owl2-rdf-based-semantics/#Appendix:_Relationship_to_the_Direct_Semantics_.28Informative.29
>
>



-- 
Dr. Birte Glimm, Room 306
Computing Laboratory
Parks Road
Oxford
OX1 3QD
United Kingdom
+44 (0)1865 283529

Received on Monday, 2 November 2009 00:14:13 UTC