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

RE: Models and their use

From: Myers, Jim <MYERSJ4@rpi.edu>
Date: Wed, 13 Jul 2011 13:53:59 -0400
Message-ID: <B7376F3FB29F7E42A510EB5026D99EF2053FFB4E@troy-be-ex2.win.rpi.edu>
To: Simon Miles <simon.miles@kcl.ac.uk>, Provenance Working Group WG <public-prov-wg@w3.org>
> 
> Jim,
> 
> > Leaving out description, which I agree is separable, where do we stand with
> thing/IVP?
> > Thing is being talked about as entity now (and was Bob for a while)?
> 
> But this is exactly the problem - from the F2F discussions, there ended up
> being three separate concepts: thing/BOB, stuff/entity, and entity state.
> Thing was not being talked about as entity, but as the "representation of
> entities with properties modelling aspects of entity states"
> 
> > And the argument is that we don't need to distinguish between the idea
> > of a provenance thing and the entity definitions in other ontologies
> 
> That is a possible consequence of the disentanglement of concepts.
> 
> > , or perhaps better - the mapping of instances of the classes 'comet' and
> 'comet-near-star' to provenance entities is not part of the model, but is an
> implementation detail?
> 
> Not sure about this... what is the relevant difference between an instance
> and an entity?

I think this is the 'thing as representation of entity' debate. I think we require that a provenance thing/entity/bob has to have an ID and a set of properties with values that can be retrieved. Instances of classes in other ontologies, or of programming language classes etc. could be richer - have behaviors, restrictions on property values, etc. I think our use cases don't require anything but ID and property/values (the 'content' as well if that's not a property), but, in analogy with the idea of process execution having a link to recipe, I could see reasons to have a link from the provenance thing/bob to the class type or full object it represents. Just like with recipe and the ability to compare what happened with what was supposed to happen, being able to compare what is asserted about things with the knowledge about how those things are supposed to be constrained/behave would be very valuable as evidence and to assess trust.
> 
> > How do we know that "Halley's Comet" and "Halley's comet near the sun"
> are related and not 'representations' of different 'stuff'? ("Kahoutek-near-
> the-sun" Is not an IVP of/not related to "Halley's comet".) What
> terms/description of 'representation' and 'stuff' is current? Do we still agree
> that there is something out there behind/beyond entities?
> 
> I don't see what "behind/beyond entities" means, and I don't remember
> agreeing with it :-). If you mean that "Halley's Comet" is behind/beyond
> "Halley's comet near the sun", then I agree there is a relation to be captured,
> and I thought IVP of did that.

I'm asking about what is being asserted with IVP of - Halley's-near-the-sun is an IVP of "Halley's" and not "Kahoutek's" because... ? We know it's not because of the cosmetic similarity of the ID strings, and, as far as we know, there are no formal relationships between the 'comet' and comet-near-the-sun' classes/ontologies (I'm assuming this is the general case). In common sense terms, we have some notion that, at least at some times, there is 'stuff' - the rocks/dust/water/etc. in "Halley's" and "Halley's-near-the-sun" that  is common to both ("comet" could mean just the nucleus and "comet-near-the-sun" might include water in the tail, so there may still only be partial overall of the set of particles in both). Stuff was aimed at that level - there's something more fundamental than the notion of comet, such as a 'set of particles/molecules' (if you believe in those), and at that level of representation Halley's and Halley's-near-the-sun have some common set members in that representation, whereas Kahoutek and Halley's-near-the-sun do not. If you believe in wave functions, string theory, etc. you may disagree what the right level for assessing commonality is. 

If we don't like 'stuff' and don't want to talk about entities as representation of /conceptualization of stuff as a way to define what IVP of means, what's the alternative being proposed?
> 
> On the other hand, if we keep "representation" as part of thing/entity's
> definition, then "Halley's comet near the sun" is not a thing/entity, only the
> "representation of Halley's comet near the sun". This complexity shouldn't
> arise coz the model should be about the world and its usage about the
> representation.

Not sure I can parse this. I would say the neither comets nor comets-near-the-sun are real or representations of each other - they are theoretical constructs that are a useful shorthand in talking about more fundamental stuff and the processes that affect it. I would like to assert an IVP relationship between things/entities in these two constructs  when I believe those entities are different ways of talking about the same stuff and when the entity on the invariant end of the relationship is more appropriate for discussing the occurrence of certain types of processes (certain process executions) because that theoretical construct it is defined in terms of results in a new entity when the process execution occurs (versus merely changing property values in the other theoretical construct).


> 
> > Where does IVP stand? What I'm picking up for the most part is that IVP
> can't be hierarchical in general, which I think has faded as we tried to talk
> about properties of one entity having to be immutable in an IVP of it.
> Technically, I'm not sure we ever said that there couldn't be properties going
> in the reverse direction (allowing A IVP of B and B IVP of A), and my original
> arguments were that invariance was relative to the processes being
> discussed which also implies IVP relationships could go in both directions
> depending on the processes you wanted to talk about. Is the current
> discussion just reaffirming that we're not talking about hierarchies of
> invariance, or is there more to the discussion about how the IVP relationship
> has to change?
> 
> As far as I can see, this is a separate point from the discussion on this thread.
> I don't think either the current definitions or my suggested alternatives
> preclude properties in the reverse direction.

OK - so no disagreement that IVP of is not hierarchical in the general case?
> 
> With regard to invariance relative to processes, I agree that my proposal does
> not capture that, though I don't think the original definitions did either. The
> intuition implied at the start of this thread was that there are many things
> which assertions could be relative to and that this was collectively the
> "perspective" of the assertion. I suggested that we simply state that any
> assertion using a
> (our) model is by necessity from some perspective, and not make it part of
> any one definition. This allows for the simplicity of definitions urged by
> Graham. Is that inadequate?

The discussion of IVP of implying a connection between mutable and invariant properties was actually an attempt to avoid talking about 'relative to a process' and I think implicitly was a way to define the perspective just in terms of the "A IVPof B" statement. That attempt might be a cause of the complexity that is being discussed.

To answer "is this adequate?", I still have to understand what concrete change this makes to the model:

You state "from some perspective" above, but I tend to think of IVP as involving a shift in perspective - is that still consistent with your proposal - that different assertions separated by IVP relationships could have different perspectives?

And is the argument then basically that any attempt to formalize the 'perspective' leads to complexity/confusion that we should avoid? If so, it sounds tempting but I would be concerned if that ends up making it impossible to assess an IVPof assertion - e.g. if I did know the set of particles in "Halley's" and "Kahoutek's-near-the-sun", can I make any assessment of whether "Kahoutek-near-the-sun" is an IVPof "Halley's" is a valid assertion (or less stringently, that I should trust other statements of the asserter)? If the idea is to drop the constraints on properties, but to preserve this level of meaning of IVPof, then I think it's useful. If we completely lose the ability to make truth/trust assessments about IVPof assertions (using knowledge external to the model as would be the case with recipe links and assessing truth/trust in process execution assertions), then I think it goes too far.

I fear we may still not be talking the same language, but let's see - hopefully we can converge.

 Jim




> 
> Thanks,
> Simon
> 
> 
> >  Thanks,
> >  Jim
> >
> >
> >
> >
> >> -----Original Message-----
> >> From: public-prov-wg-request@w3.org [mailto:public-prov-wg-
> >> request@w3.org] On Behalf Of Simon Miles
> >> Sent: Wednesday, July 13, 2011 8:36 AM
> >> To: Provenance Working Group WG
> >> Subject: Re: Models and their use
> >>
> >> Jim, Graham,
> >>
> >> Reading through your comments, I'm pretty much in agreement with
> them.
> >> The discrepancy between my view and Jim's is, I think, just a matter
> >> of what we assume is being defined at the moment (i.e. by entity/thing).
> >>
> >> As in Graham's last mail, I am assuming that we are first defining
> >> any
> >> (contextualised) entity ((1) in Graham's classification).
> >>
> >> I fully agree with you that assertion/observation/description (2) is
> >> also an important and distinct concept. However, I see it as
> >> secondary to the idea of entity in the model and, as Graham says, a
> >> description is itself a kind of entity. This may be comparable to
> >> agents or process executions, i.e. agents, executions and
> >> descriptions are all particularly important (for provenance) subclasses of
> entity.
> >>
> >> And I think this difference in our views about what was being defined
> >> by "thing" nicely illustrates the original point I was trying to make
> >> about the conflation in the definition :-)
> >>
> >> Thanks,
> >> Simon
> >>
> >> On 13 July 2011 11:01, Graham Klyne <GK@ninebynine.org> wrote:
> >> > Jim,
> >> >
> >> > I've been thinking about your comments, and have come to a view
> >> > that there are two things going on here, which are not being
> >> > clearly
> >> distinguished:
> >> >
> >> > 1. an entity constrained to some context
> >> >
> >> > 2. observations or descriptions of an entity
> >> >
> >> > which when combined can model observations/descriptions of an
> >> > entity constrained to some context.  I have been focusing on (1),
> >> > with the expectation that (2) would be dealt with separately in the
> >> > model, where provenance is a kind of description.  If I understand
> >> > correctly, you are particularly concerned to distinguish between
> >> > entity and description.  It seems to me that "bob" has been adopted
> >> > variously to fulfil
> >> both these distinctions.
> >> >
> >> > ...
> >> >
> >> > Example: suppose we're interested in Halley's comet, and in
> >> > particular in the periods when it is close to the sun or visible
> >> > from earth.  So we have two
> >> > concepts:  Halley's comet and Halleys comet when close to Earth.
> >> > To my mind, these are both *entities* (sensu F2F1).  Any
> >> > description that is true of Halley's comet generally should also be
> >> > true of it when close to Earth, but there are many other assertions
> >> > are true the constrained Halley's comet that are probably not true
> >> > at all times (e.g. the appearance of a tail due to expulsion of gas
> >> > and dust caused by
> >> solar heating).
> >> >
> >> > The page http://en.wikipedia.org/wiki/Halley's_Comet is a
> >> > *description* of Halley's comet.  Being a Wikipedia page, it
> >> > probably changes over time.  So we may wish to discuss it as it
> >> > exists in a particular period of time.  Today, for example, it
> >> > contains a list of
> >> > 95 references indicating sources of information used in the page.
> >> > Thus we may expect this page to be a description of Halley's comet
> >> > for its entire lifetime, but needs to be constrained to make
> >> > statements about the number of references it contains.  In this
> >> > respect, the page is
> >> both an "entity" and a "description" of an entity.
> >> >
> >> > The page at http://en.wikipedia.org/wiki/Halley's_Comet also has a
> >> > brief description of a Cuneiform-inscribed clay tablet which is
> >> > both an entity in its own right, with its own description and
> >> > provenance,
> >> > *and* a record of observation of Halley's comet.  I don't think we
> >> > can easily disjoin the class of descriptions from the class of entities.
> >> >
> >> > Thus, I perceive that the notion of constraint so that some aspects
> >> > of a constrained entity are invariant needs to be dealt with
> >> > separately from the notion of description, which often (but not
> >> > always, I think) applies to a constrained form of some entity.
> >> >
> >> > Does this make any sense?
> >> >
> >> > #g
> >> > --
> >> >
> >> >
> >> > Jim McCusker wrote:
> >> >> My issue is more with Simon's desire to conflate things with the
> >> >> descriptions of those things. We need to be able to say "x, as
> >> >> described by y", which is a separate issue from relating "x1" as
> >> >> an invariant view of "x2".
> >> >>
> >> >> "x as described by y" can be done in semweb using a URI for X and
> >> >> the URI for a named graph or graph literal for y, assuming that
> >> >> there are assertions in y about x. Another alternative is that y
> >> >> can be discussed as an information artifact, as in the Information
> >> >> Artifact Ontology (http://code.google.com/p/information-artifact-
> ontology).
> >> >> Since x changes over time, we need something that nails down what
> >> >> was actually "seen", or at least, claimed when the agent
> >> >> identified x in its context.
> >> >>
> >> >> A BOB is the y in the above paragraph, and I think that the fact
> >> >> that it's something that's describing an x, it must have the extra
> >> >> qualification in place in its name.
> >> >>
> >> >> Jim
> >> >>
> >> >> On Tue, Jul 12, 2011 at 3:27 PM, Graham Klyne <GK@ninebynine.org>
> >> wrote:
> >> >>> [Off-list]
> >> >>>
> >> >>> Jim,
> >> >>>
> >> >>> I had some small reservations about Simon's definition, but I
> >> >>> felt that it was such a big improvement over what we currently
> >> >>> have that I didn't want to muddy the waters just yet by adding
> >> >>> qualifications to my support :)
> >> >>>
> >> >>> (Specifically, I would remove "and something more is invariant
> >> >>> about B" from the efinition of invariant view, so that anything
> >> >>> can be an invariant view of itself (or not excluded from so
> >> >>> being) - which I think is one of the concerns you raised.)
> >> >>>
> >> >>> Apart from that, I think there is an aspect of an invariant view
> >> >>> that is in some sense fundamentally subsumptive -- there is a
> >> >>> distinct sense that A and B are generally the same, except that
> >> >>> one may
> >> be more constrained.
> >> >>>
> >> >>> But, more importantly, I think we need to be looking to say less,
> >> >>> not
> >> more.
> >> >>> I feel that Simon's definition captures close to what we need to
> >> >>> say without adding too much more.
> >> >>>
> >> >>> In this, I'm arguing for the minimum useful semantics - it's
> >> >>> easier to add (or layer) constraints later than to remove them
> >> >>> from an established defintion.  By providing a little as we can
> >> >>> for people to disagree with, I think we maximize the potential
> >> >>> for take-up of the
> >> WG outputs.
> >> >>>
> >> >>> #g
> >> >>> --
> >> >>>
> >> >>>
> >> >>> Jim McCusker wrote:
> >> >>>> On Tue, Jul 12, 2011 at 12:02 PM, Graham Klyne
> >> >>>> <GK@ninebynine.org>
> >> wrote:
> >> >>>>> Simon Miles wrote:
> >> >>>>>> To understand the consequences of the above points, I suggest
> >> >>>>>> alternative definitions at the link below:
> >> >>>>>>
> >> >>>>>>
> >> >>>>>>
> >> http://www.w3.org/2011/prov/wiki/Talk:F2F1ConceptDefinitions#Enti
> >> >>>>>> ty_and_IVP_of
> >> >>>>> +1
> >> >>>>>
> >> >>>>> I think this is a big improvement over what we have.
> >> >>>> I like the Entity definition, but I'm not sure how we then go
> >> >>>> about qualifying assertions about Entities. We need a way of
> >> >>>> making those assertions (which is what BOBs were for) and a way
> >> >>>> of relating Entities that are the same, even if they aren't
> >> >>>> mathematically the same (different state, different aspect,
> >> >>>> etc.). IVP of as it's defined there is not quite enough, since
> >> >>>> it only allows for relations between entities that have
> >> >>>> subsumptive (a is IVP of b, therefore a has all the states of b plus
> some).
> >> >>>>
> >> >>>> Jim
> >> >>>> --
> >> >>>> Jim McCusker
> >> >>>> Programmer Analyst
> >> >>>> Krauthammer Lab, Pathology Informatics Yale School of Medicine
> >> >>>> james.mccusker@yale.edu | (203) 785-6330
> >> >>>> http://krauthammerlab.med.yale.edu
> >> >>>>
> >> >>>> PhD Student
> >> >>>> Tetherless World Constellation
> >> >>>> Rensselaer Polytechnic Institute mccusj@cs.rpi.edu
> >> >>>> http://tw.rpi.edu
> >> >>>>
> >> >>>
> >> >>>
> >> >>
> >> >>
> >> >>
> >> >
> >> >
> >> >
> >>
> __________________________________________________________
> >> ____________
> >> > This email has been scanned by the MessageLabs Email Security System.
> >> > For more information please visit http://www.messagelabs.com/email
> >> >
> >>
> __________________________________________________________
> >> ____________
> >> >
> >>
> >>
> >>
> >> --
> >> Dr Simon Miles
> >> Lecturer, Department of Informatics
> >> Kings College London, WC2R 2LS, UK
> >> +44 (0)20 7848 1166
> >
> >
> >
> __________________________________________________________
> ____________
> > This email has been scanned by the MessageLabs Email Security System.
> > For more information please visit http://www.messagelabs.com/email
> >
> __________________________________________________________
> ____________
> >
> 
> 
> 
> --
> Dr Simon Miles
> Lecturer, Department of Informatics
> Kings College London, WC2R 2LS, UK
> +44 (0)20 7848 1166
Received on Wednesday, 13 July 2011 17:55:31 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 13:06:37 GMT