Re: Linking Across Provenance Bundles - comments

On 30/11/2012 17:30, Luc Moreau wrote:
> Hi Graham, all,
>
> The staged document can be found at:
> http://dvcs.w3.org/hg/prov/raw-file/default/links/releases/WD-prov-links-20121211/Overview.html
>
>
> Let me know if you find any bug.
>
> I have added links to the dm/constraints sections that use the term 'aspect' and
> I have also
> added its informalmeaning.
>
> I didn't understand the conflict you suggested. Can you clarify?

The conflict I see here is the implication that by simply making assertions 
about an entity you are somehow defining a new and changed entity.  If I say 
that the provenance linking document is authored by you and Tim, am I thereby 
creating a new entity based on that document?  I don't think so, but the text 
about defining a new entity for which the bundle is a new fixed aspect suggests 
that this is what is happening.

> Finally, the notions of observer, perspective, context are a very significant
> departure from what
> we currently have, and the dm in general.  I don't feel comfortable with them in
> the document.

To my mind, it's only by considering the context of an entity described by a 
bundle that the specialization-by-mention makes any sense at all.

The elephant in the room here, which I think we have been dancing around, is 
that the mention construct introduces a context for the statements in a bundle. 
  It's only through restricting the observations in a bundle to this 
(existential) context that I can make any sense of the idea that the bundle can 
be used to designate a specialization of an entity.

Working from the implications of mention that Tim described to me at the F2F, I 
could only make sense of this by associating the bundle observations with a 
specialization, which led to the text that the group (reasonably) felt had no 
greater clarity.  I think this will remain a muddled concept if it is not 
grounded in some notion of context or observation perspective.

...

As a NOTE, I don't propose to make a big deal of this problem, but without 
greater clarity I fear that it will be of little practical value in creating a 
common view of what nature of specialization is introduced by a mention 
statement.  Ultimately, what I think it needs is a formal semantics.

#g
--

> On 11/29/2012 06:15 PM, Graham Klyne wrote:
>> http://dvcs.w3.org/hg/prov/raw-file/default/links/prov-links.html
>>
>> My comments are focused on Section 2, Conceptual Definition of Mention
>>
>> [[
>> An entity e1 may be mentioned in a bundle b, which contains some descriptions
>> about this entity e1: how e1 was generated and used, which activities e1 is
>> involved with, the agents e1 is attributed to, etc. Other bundles may contain
>> other descriptions about the same entity e1. Some applications may want to
>> augment the descriptions of entity e1 found in bundle b with other
>> information. To this end, PROV allows a new entity e2 to be created and
>> defined as a specialization of the preceding entity e1, and which presents at
>> least an additional aspect: the bundle b containing some descriptions of e1.
>> With this relation, applications that process e2 can know that the attributes
>> of e2 may have been computed according to the descriptions of e1 in b.
>> ]]
>>
>> "Some applications may want to augment the descriptions of entity e1 found in
>> bundle b with other information" appears to be in conflict with the earlier
>> statement "Other bundles may contain other descriptions about the *same*
>> entity" (my emphasis).
>>
>> [[
>> The mention ◊ of an entity in a bundle (containing a description of this
>> entity) is another entity that is a specialization of the former and that
>> presents the bundle as a further additional aspect.
>>
>> An entity is interpreted with respect to a bundle's description in a domain
>> specific manner. The mention of this entity with respect to this bundle offers
>> the opportunity to specialize it according to some domain-specific
>> interpretation.
>> A mention of an entity in a bundle results in a specialization of this entity
>> with extra fixed aspects, including the bundle that it is described in.
>> ]]
>>
>> I have two concerns here:
>>
>> This definition leans heavily on the notion of "fixed aspects", but this term
>> is not actually defined anywhere.
>>
>> Philosophically, I have difficulties with the implication that simply by being
>> mentioned in a bundle somehow changes the entity itself. Suppose you are
>> looking at an elephant, and writing down a description of that elephant.  Then
>> I come along and look at the elephant from a different perspective, and write
>> down a description of what I see.  We may put down different descriptions, but
>> we are describing the *same* elephant. The fact of me doing that doesn't
>> change the elephant in any way, nor does it affect the validity of your
>> description concerning the elephant.  Yet the idea that an observation is a
>> new "fixed aspect" seems to suggest that this is act of observation makes it a
>> different elephant.  (Not allowing Heisenberg/observer effects here.)
>>
>> What I think is missing from your description is some notion of the
>> perspective ("observation context") from which the descriptions in a bundle
>> are derived.
>>
>> So here's my attempt to try and capture something of this:
>>
>> [[
>> Statements about an entity are based on information available from some
>> perspective, and the extent of information available may be affected by the
>> context of an observer that records it.  The mention construct provides a way
>> to make additional statements about the entity from the same context or
>> perspective as other statements in a bundle.  For example, one bundle may
>> contain statements about a web page accessed from a network connection in USA
>> which is presented containing comments authored by Americans, and another may
>> contain statements about the same page accessed at the same time from China,
>> which may contain a different set of comments.  Deducible provenance about the
>> contributors to this page may thus vary based on where it is accessed from.
>>
>> The mention construct provides a way to associate this additional contextual
>> information with an entity, such as the location from which it has been
>> accessed when making some provenance assertions.
>> ]]
>>
>> #g
>> --
>>
>>
>

Received on Saturday, 1 December 2012 11:53:32 UTC