W3C home > Mailing lists > Public > public-prov-wg@w3.org > October 2011

Re: PROV-ISSUE-104 (time-class): How to relate start/end time to PE, use, generation, etc [Formal Model]

From: Stian Soiland-Reyes <soiland-reyes@cs.manchester.ac.uk>
Date: Fri, 14 Oct 2011 10:14:53 +0100
Message-ID: <CAPRnXtmDF09JNChN+eSHvQ13sZZDGNUz7z9n+-v4OJpL+=VhGA@mail.gmail.com>
To: Paul Groth <p.t.groth@vu.nl>
Cc: Provenance Working Group WG <public-prov-wg@w3.org>
The DM constraint generation-unicy (or so) says that if in a single account
asserts:

wasGeneratedBy(e, pe1, q1)
wasGeneratedBy(e, p2, q2)

then:
p1==p2 and q1==q2

I assume it follows that also the generation times would equal, but that
should probably be added to the constraint. Therefore a single functional
prov:wasGeneratedAt should be OK.

Now it is open for discussion if this constrain interferes with subprocesses
where an entity can be seen as "generated two places at once", or if that
should be solved by two (potentially roled) entities or two accounts.

You can do the same EntityInRole construct for prov:wasGeneratedBy.
Syntactically each of these are new entities, can all have prov:assumedBy
:e1 but  different prov:assumedRoles, prov:wasGeneratedAt (and in theory
prov:wasGeneratedBy).

By the constraint they should then be interpreted as the same generation
process and qualifiers (including role). You can think of it as an inference
rule when an entity in role states its generation:

x prov:assumedBy e1
  prov:wasGeneratedBy :p1
  prov:wasGeneratesAt :t1
-->
x prov:wasComplementOf e1
e1 prov:wasGeneratesBy :p1
  prov:wasGeneratesAt :t1

ie they overlap, but have common start time. Again stating entity start/end
time (characterisation unrerval) would make this easier.

At the same time I feel that this pushes EntityInRole to its limits. That's
good, because we need to see if it is up to the job.

Tim, Satya, any comments?
On Oct 14, 2011 4:50 AM, "Paul Groth" <p.t.groth@vu.nl> wrote:

> This is a good start:
>
> But how is for example wasGeneratedAt associated with a particular process
> execution generated pair. Or do we assume that two such edges shouldn't
> occur?
>
> I guess this done through the blank node in the used example is that
> correct?
>
> Thanks for this
> Paul
>
> On Oct 14, 2011, at 1:30, Stian Soiland-Reyes <
> soiland-reyes@cs.manchester.ac.uk> wrote:
>
> >> On Thu, Oct 13, 2011 at 17:49, Paul Groth <p.t.groth@vu.nl> wrote:
> >>
> >>> I was wondering if examples of time could be put in the PROV-O Crime
> >>> Example. Like they are in PROV-DM.
> >>>
> >>> It would be cool to see how we can put times against:
> >>>
> >>> used,
> >>> wasGeneratedBy
> >>> Provenance Execution.
> >
> > I threw together one suggestion on how to do it using a few new
> properties:
> >
> > http://dvcs.w3.org/hg/prov/file/tip/ontology/components/Time.ttl
> >
> > declares:
> >
> >  prov:startedAt
> >  prov:endedAt
> >  prov:assumedRoleAt
> >  prov:wasGeneratedAt
> >
> >
> > For example, from
> >
> http://dvcs.w3.org/hg/prov/file/285b39f7ea5f/ontology/components/Time/example-1.ttl
> >
> > # start/end
> > :pe1 a prov:ProcessExecution ;
> >    prov:startedAt :tA ;
> >    prov:endedAt :tB .
> > :tA a prov:Time .
> > :tB a prov:Time .
> >
> >
> > # generation time
> > :e1 a prov:Entity ;
> >    prov:wasGeneratedBy :pe1 ;
> >    prov:wasGeneratedAt :tC .
> > :tC a prov:Time .
> >
> >
> > # used time
> > :pe2 a prov:ProcessExecution ;
> >    prov:used [
> >        a prov:EntityInRole;
> >        prov:assumedBy :e1;
> >        prov:assumedRoleAt :tD
> >    ] .
> > :tD a prov:Time .
> >
> >
> > # controlled time
> > :a1 a prov:Agent .
> > :tE a prov:Time .
> > :pe2 prov:wasControlledBy [
> >        a prov:EntityInRole, prov:Agent;
> >        prov:assumedBy :a1;
> >        prov:assumedRoleAt :tE;
> >    ] .
> >
> >
> > You can from this of course also specify generation using EntityInRole
> > and assumedRoleAt - but as an entity can only be generated once
> > (right?) we might as well just say prov:wasGeneratedAt.
> >
> >
> >
> > Note to self:
> > Allow intervals for role/use/control? - (prov:assumedRoleStart and
> > prov:assumedRoleEnd)?
> > Sounds like special cases of characterisation intervals..
> > prov:startedAt and prov:endedAt could also be used with prov:Entity ?
> >
> >
> >
> > I looked at the crime file example, and did:
> >
> >
> http://dvcs.w3.org/hg/prov/file/285b39f7ea5f/ontology/components/Time/example-2-crimefile.ttl
> >
> > .. but it does not really say much about time at all:
> >
> >
> > #processExecution(pe1,add-crime-in-london,t+1,,[])
> > :pe1 a prov:ProcessExecution, cf:FileAppending ;
> >    prov:startedAt :t1 .
> >
> > #processExecution(pe2,email,t+2,,[])
> > :pe2 a prov:ProcessExecution, cf:Emailing ;
> >    prov:startedAt :t2 .
> >
> >
> >
> > So here's a practical example of how to do an extension, for instance
> > using OWL Time:
> >
> >
> http://dvcs.w3.org/hg/prov/file/285b39f7ea5f/ontology/components/Time/example-3-extension.ttl
> >
> > Including the simplest case without using the subclasses:
> >
> > # Lazy straight forward version - no declared subclasses
> > :pe0 a :ProcessExecution ;
> >    prov:startedAt [
> >        # a prov:Time, time:Instant ;
> >        time:inXSDDateTime "2011-12-24T14:24:05Z"
> >    ] ;
> >    prov:endedAt [
> >        time:inXSDDateTime "2011-12-25T01:41:54Z"
> >    ] .
> >
> >
> >
> >
> > --
> > Stian Soiland-Reyes, myGrid team
> > School of Computer Science
> > The University of Manchester
>
Received on Friday, 14 October 2011 09:15:23 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 13:06:44 GMT