Hi Nina,

On Wed, 2010-10-06 at 18:23 +0300, Nina Jeliazkova wrote:

> Hi Pantelis,
> 
> Just to add - if I understood right, it might require some work to parse non
> - OntModel into ToxOtis internal objects, but what I would be interested is
> the time to parse the incoming stream into Jena model only, ignoring further
> transformations for the time being.


We've put this in high priority so that in case there's a significant
difference between OntModel and other Model implementations, we'll
refactor a bit the source code. So we'll have the related results by
tomorrow...

Pantelis

> 
> Do you think this kind of measurement is feasible?
> Nina
> 
> On Wed, Oct 6, 2010 at 6:17 PM, Nina Jeliazkova
> <jeliazkova.nina at gmail.com>wrote:
> 
> > Hi Pantelis,
> >
> > There is no problem of parsing OWL-DL with Jena models (recall OWL-DL is a
> > valid RDF) , different than OntModel.
> >
> > You will only loose some convenient methods and classes in Jena, but you
> > can still check if a class has certain rdf:type or certain properties.  I
> > agree the client code might became less elegant , but performance gain may
> > worth it.
> >
> >
> > On Wed, Oct 6, 2010 at 6:08 PM, chung <chvng at mail.ntua.gr> wrote:
> >
> >> Hi Nina, Egon,
> >>   That would be an interesting experiment in general but there are some
> >> technical problems related to ToxOtis and to the current format of our
> >> RDF documents. First of all, a common RDF specification has been chosen,
> >> that is OWL-DL and we have lots of times striven to stick to it. ToxOtis
> >> produces OWL-DL compliant RDF representations for every OpenTox entity
> >> (methods asIndividual(OntModel):Individual and asOntModel():OntModel ).
> >> OntModel and OntClass have been chosen as the prime Jena objects to work
> >> with in ToxOtis and in YAQP, our web application also. Apart from the
> >> fact that it would be hard to change that, I also reckon we shouldn't
> >> for various reasons I'll try to summarize.
> >>
> >
> >
> >   OntModel is an interface that offers lots of functionalities such as
> >> creation of Annotation, Datatype and Object Properties that are missing
> >> from the interface Model where one can just use m.createProperty(String)
> >> to create an untyped property. Additionally, OntModel is tightly
> >> connected to OntClass, an interface used to describe the ontological
> >> classes to which instances are binded. So if we need to produce OWL-DL
> >> compliant representations we either need an implementation of OntModel
> >> and OntClass provided by Jena or one of ours. My opinion is that the
> >> adoption of OWL-DL against OWL-Full or OWL-Lite does not increase
> >> significantly the size of the representation. What is more, OWL-DL
> >> guarantees computability and decidability and does not oblige users to
> >> use some inference engine that imposes overhead so it would be better to
> >> cling to it both for the server and from the client side.
> >>   As you already know, using OntModel we have performed measurements to
> >> compare between different specifications and it was shown that OWL-DL
> >> performs better for parsing documents compared to OWL and OWL-Lite. What
> >> might be of interest is to reveal the impact of the cache size (for in
> >> memory triple storage) on the performance of RDF parse/serialization
> >> procedure and on the allocation of resources.
> >>
> >> Best Regards,
> >> Pantelis
> >>
> >> On Wed, 2010-10-06 at 15:12 +0200, Egon Willighagen wrote:
> >>
> >> > Pantelis,
> >> >
> >> > On Wed, Oct 6, 2010 at 2:58 PM, Nina Jeliazkova
> >> > <jeliazkova.nina at gmail.com> wrote:
> >> > >> Would it be hard for you to switch to "plain" RDF (i.e.  omit all
> >> > >> RDF:type statements) and compare computation times with OWL
> >> > >> representations.
> >> > >
> >> > > From dataset service point of view it is practically impossible,
> >> without
> >> > > breaking everything,  but form client point of view one can try using
> >> Jena
> >> > > models, different than OntModel.
> >> >
> >> > what happens if you use either of these:
> >> >
> >> > ModelFactory.createDefaultModel()
> >> > ModelFactory.createNonreifyingModel()
> >> >
> >> > ?
> >>
> >>
> >> If you think this is necessary, we can set up an experiment without
> >> ToxOtis...
> >>
> >
> >
> > Yes, please (unless it is a month work...)
> >
> > Best regards,
> > Nina
> >
> >
> >
> >
> >>
> >> >
> >> > Egon
> >> >
> >>
> >>
> >> _______________________________________________
> >> Development mailing list
> >> Development at opentox.org
> >> http://www.opentox.org/mailman/listinfo/development
> >>
> >
> >
> >
> _______________________________________________
> Development mailing list
> Development at opentox.org
> http://www.opentox.org/mailman/listinfo/development
>