Re: PROV-DM derivation concerns arising from my primer review

On Fri, Nov 25, 2011 at 12:56, Graham Klyne <Graham.Klyne@zoo.ox.ac.uk> wrote:

> You make a good case here for something stronger that the weakest form of
> derivation, by being clear that the precise meaning is in the beief of the
> asserter.  It's a different modality, maybe "believedToBeDerivedFrom"?

I don't think we need the "believed to" bit - the asserter believes
all the other things. The entity was believed to been generated such
as such, the process execution was believed to have used such and
such. It is not up to us to limit or constraint how the asserter came
to the conclusion - but we can easily recognise that it is possible to
conclude that something was derived from something else, even without
knowing the exact processes involved.


>> d) Something I didn't think of - perhaps he made a subproperty that is
>> stronger than possiblyDerivedFrom but not as strong as
>> necessarilyDerivedFrom
> I'm not sure about (d) if the "necessarilyDerivedFrom" is in the belief of
> an asserter, and has no formal semantics, how can we say it's stronger than
> "possiblyDerivedFrom" (other than as informal claim)?

I did not argue that necessarilyDerivedFrom should have any formal
semantics, I argued it is a subproperty of possiblyDerivedFrom (which
then is not a good name, obviously) - saying that not just was there a
(chain of) use-generation-control that links the two entities, but
yes, the final entity is "truly" derived from the first one. We
recognise this "true derivation" as richer than the "was just involved
somewhat" derivation of "possiblyDerivedFrom".

So you've got two conclusions to draw from a necessarilyDerivedFrom,
a) there was such a chain of use-generation-control as with
possiblyDerivedFrom, b) the asserter felt this is strongly a "true"
derivation, and so has (also) asserted it as such explicitly.


If you form a subproperty of possiblyDerivedFrom, say
ex:wasAResponseTo - then you can form a domain-specific derivation
that comes with the added understanding that there was a such chain of
process executions. (but we might not know all the details of that
chain).

However it is not stating that the first entity (necessarily) has
directly contributed to the second entity. The distinction from being
a subproperty of "possiblyDerivedFrom" or "necessarilyDerivedFrom"
would be not very noticeable for consumers which understand the
extension, but is there as a helpful hint to general provenance
exchange - which is what PROV is all about.

What "stronger" above means would just be an informal thing that we
can't detail, the old notion that some attributes have to be decided
by the other attributes is a bit too specific, but highlights the
level of "affecting" we are talking about, and we can easily show this
by example, but say that it is up to the asserter and domain to
ultimately decide. If in doubt, they can always use the "weaker"
possiblyDerivedFrom.


-- 
Stian Soiland-Reyes, myGrid team
School of Computer Science
The University of Manchester

Received on Monday, 28 November 2011 13:50:04 UTC