Re: [Owlim-discussion] reasoning context separation in OWLIM?

[Ivan Peikov]

Hi Jacek,

You are perfectly correct in your conclusions -- inferred statements end up in 
the default context and are kind of shared between all contexts. Also, there 
is now way of configuring those to be stored in their source context(s). 

When context separation is required, the usual workaround is to tie your graph 
patterns with the desired context and only allow inferred triples to be 
retrieved from the default context, e.g.

SELECT * WHERE {
  GRAPH <http://context/graph> {
    ?A some:predicate ?B
  }
  ?A rdf:type ?C
}

In this example the some:predicate triple would be known to exists in some 
particular context http://context/graph and the rdf:type statement will be 
inferred in the default context but as A would be bound to the same A from 
the desired context we would be sure that we don't get triples from undesired 
contexts.

Is this workaround relevant to your case?


Cheers,
Ivan

On Tuesday 18 May 2010 23:31:50 J.Kopecky wrote:
> Dear OWLIM list members,
>
> I'm new here as a recent user of OWLIM in the iServe project [1]. I have
> two questions, the second will come in the next mail.
>
> We use OWLIM with the following configuration:
>
> [] a rep:Repository ;
>    rep:repositoryImpl [
>       rep:repositoryType "openrdf:SailRepository" ;
>       sr:sailImpl [
>          sail:sailType "swiftowlim:Sail" ;
>          owlim:ruleset "owl-max" ;
>          owlim:partialRDFS "true" ;
>          owlim:noPersist "true" ;
>          owlim:storage-folder "main-data-storage" ;
>          owlim:base-URL
> "http://www.ip-super.org/ontologies/execution-history#"; ;
> owlim:new-triples-file "new-triples-file.nt" ;
>          owlim:entity-index-size "200000" ;
>          owlim:jobsize "200" ;
>          owlim:repository-type "in-memory-repository" ;
>          owlim:defaultNS
> "http://www.ip-super.org/ontologies/execution-history#"; ]
>    ].
>
> The repository does RDFS materialization, but all the materialized triples
> are apparently in the default context (if this is the right term - I view
> the repository through the Sesame OpenRDF Workbench). As far as I
> understand it, this means that we cannot separate different contexts in
> terms of inferred triples - is this correct? If one context implies the
> triple <A,rdf:type,B>, we cannot query another context for the instances of
> B without also getting A, right?
>
> Is there a configuration of OWLIM that does context separation for
> reasoning? If not, what is the recommended workaround?
>
> Thanks,
> Jacek Kopecky
> Knowledge Media Institute, The Open University, UK
>
>
> [1] http://iserve.kmi.open.ac.uk/


_______________________________________________
OWLIM-discussion mailing list
OWLIM-discussion@ontotext.com
http://ontotext.com/mailman/listinfo/owlim-discussion