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

Re: PROV-ISSUE-85 (What-is-Entity): Definition of Entity is confusing, maybe over-complex [Conceptual Model]

From: Graham Klyne <GK@ninebynine.org>
Date: Tue, 06 Sep 2011 12:40:15 +0100
Message-ID: <4E66069F.1040403@ninebynine.org>
To: Luc Moreau <L.Moreau@ecs.soton.ac.uk>
CC: public-prov-wg@w3.org
Hi Luc,

On 06/09/2011 11:09, Luc Moreau wrote:
> Hi Graham,
>
> If i recall correctly, you commented on "represents", and this was replaced by
> "is representation of".

To me, those phrases are practically synonymous in the contexts we've been 
discussing.

> I am not trying to play with words here, I genuinely feel that for people
> working in knowledge representation/data modelling,
> constructing representations of the world is a well understood notion.

(1) Yes, we commonly talk about "constructing representations of the world", 
"knowledge representation", etc. in general terms.  But when, for example, we 
were discussing RDF semantics, this word very rarely came up in a technical 
context.  I have found that some of your uses of this word actually conflict 
with what I expect it to mean in the context of use.  I.e. when you say that an 
expression that makes an assertion about an object is a representation of that 
object, that jars with my understanding of "representation".  If I say "my cat 
is black", I don't think that sentence is a representation of my cat.  (Though, 
using a sense of "representation" quite different to that implied by 
"representations of the world" or "knowledge representation", it might be a 
"representation about my cat".)

And what *does* it mean to "assert a representation"?

(2) Even if what you say is true, I don't think that's the only audience for 
this document.  I'm not sure if I count as someone who "works in knowledge 
representation/data modelling", but I've been having problems following your 
usage, and I'm sure that there will be many developers with less exposure than I 
to this field who will need to understand this.

> So, I don't see this word as problematic in this context if
> - it is used only with this meaning
> - is introduced up front to mean this.
>
> Now assuming that this still does not satisfy you, the choice of 'assertion' is
> problematic. Because sometimes,
> entities (or process executions or agents) are not asserted but inferred. It
> would really feel strange to
> write that an assertion is inferred ...

Less odd to me than the notion of "asserting a representation".  And I thought 
it had been said elsewhere that we are defining an assertion language.

But other words may be possible and less contentious: expression, description, 
predicate, statement come to mind.

#g
--

> On 06/09/11 06:42, Graham Klyne wrote:
>> Luc,
>>
>> I mentioned previously that "representation" is an imprecise word with
>> different possible interpretations. I don't think this is a good choice of
>> focus for your model description.
>>
>> Better, I think, would be to focus on "assertions" about the world, especially
>> in light of your point 2 below. To say that a "representation" can be
>> "asserted" doesn't make sense to me.
>>
>> ...
>>
>> Apart from that, then talking about data model constructs seems a reasonable
>> way to go. In a different message, I said that I thought the treatment of the
>> abstract syntax wan not helping. Reflecting further on this, I see two ways to
>> go:
>>
>> (1) drop the abstract syntax entirely, and build a model around RDF abstract
>> syntax (i.e. binary predicates, with named graphs for accounts)
>>
>> or
>>
>> (2) put the provenance abstract syntax at the heart of the model document, and
>> effectively make it all about defining the abstract syntax. Describe
>> everything in terms of PASN constructs, and focus on explaining the meaning of
>> those constructs. Don't relegate it to an appendix.
>>
>> I suspect the latter is closer to the direction you would prefer to take :)
>>
>> #g
>> --
>>
>>
>> If you are going to
>>
>> On 05/09/2011 22:22, Luc Moreau wrote:
>>> Hi Graham,
>>>
>>> Actually, this whole thread of discussion, and separate emails with Paolo,
>>> make me think that it's more appropriate:
>>>
>>> 1. To present the data model as a representation of the world.
>>> Hence, an entity is a *representation* of an identifiable characterized thing,
>>> or a process execution is a *representation* of an activity.
>>>
>>>
>>> 2. To explain that representations of the world can be asserted by asserters,
>>> and what this means in terms
>>> of existence of stuff in the world (from the asserter's viewpoint).
>>> Note that the model also allows for some of the representations to be inferred
>>> (as opposed to asserted).
>>> For instance, agents can be inferred or asserted.
>>>
>>> Hence, for uniformity, I think it's more appropriate to talk about constructs of
>>> a data model.
>>> Where it is clear that a construct can only be asserted, then, use the term
>>> assertion.
>>>
>>> Further comment interleaved.
>>>
>>> On 05/09/11 21:36, Graham Klyne wrote:
>>>> Luc,
>>>>
>>>> I think it's better, but:
>>>> (a) I still think the term "Entity" doesn't quite reflect what is being
>>>> defined, and
>>>> (b) I still think the first sentence doesn't really tackle what an Entity
>>>> [Assertion] actually *is* - your more oblique approach via "representation"
>>>> leaves me, as a reader, guessing at what it is you really mean to convey.
>>>>
>>>> I still think that starting out with something like:
>>>>
>>>> "A (BOB) is an assertion about an identifiable characterized thing" (if you'll
>>>> excuse the resurrection of "BOB" here) gets the key information in front of
>>>> the reader in a way that is less easily overlooked. Subsequent text can
>>>> explain in more detail, as you do, the details of what this actually means.
>>>
>>> I think this is very misleading. A Bob (hey, long time no see!) is not an
>>> arbitrary assertion *ABOUT* an identifiable characterized thing.
>>> In fact, when you assert a Bob, you assert the *existence* of an identifiable
>>> characterized thing (... with attributes ... over ... interval, etc).
>>>
>>> Cheers,
>>> Luc
>>>
>>>
>>>>
>>>> #g
>>>> --
>>>>
>>>>
>>>> On 05/09/2011 15:53, Luc Moreau wrote:
>>>>> Hi Graham, Jim, and Simon,
>>>>>
>>>>> Following the discussion this WE, Paolo and I have revised the definition of
>>>>> entity.
>>>>> Before editing the document, we would like to get your feedback.
>>>>>
>>>>> General assumption (to appear in section 4): in the real world, we find:
>>>>> - identifiable characterized things, their situation in the world
>>>>> - activities
>>>>> - events
>>>>>
>>>>> Cheers,
>>>>> Luc
>>>>>
>>>>>
>>>>>
>>>>> -----
>>>>> Revised section 5.1
>>>>> -------
>>>>>
>>>>>
>>>>> In PIDM, an entity construct is a representation of an identifiable
>>>>> characterized thing.
>>>>>
>>>>> An instance of an entity construct, expressed as entity(id, [ attr:
>>>>> val, ...]) in the Provenance Abstract Syntax Notation:
>>>>> - contains an identifier id, denoting a characterized thing
>>>>> - contains a set of attribute-value pairs [ attr: val, ...], representing
>>>>> this characterized thing's situation in the world.
>>>>>
>>>>> The assertion of an instance of an entity construct , entity(id, [ attr: val,
>>>>> ...]), states, from a given asserter's viewpoint, the existence of an
>>>>> identifiable characterized thing, whose situation in the world is
>>>>> represented by
>>>>> the attribute-value pairs, which remain unchanged during a characterization
>>>>> interval, i.e. a continuous interval between two events in the world (which
>>>>> may
>>>>> collapse into a single instant).
>>>>>
>>>>> Example: <same example>
>>>>> ... states the existence of a thing of type File and location
>>>>> /shared/crime.txt,
>>>>> and creator alice, denoted by identifier e0, during some characterization
>>>>> interval.
>>>>>
>>>>> Further properties:
>>>>> - If an asserter wishes to characterize a thing with same attribute-value
>>>>> pairs
>>>>> over several intervals, then they are required to assert multiple entity
>>>>> assertions, each with its own identifier.
>>>>>
>>>>> - There is no assumption that the set of attributes is complete and that the
>>>>> attributes are independent/orthogonal of each other.
>>>>>
>>>>> Cheers,
>>>>> Luc
>>>>>
>>>>>
>>>>> On 09/01/2011 05:32 PM, Provenance Working Group Issue Tracker wrote:
>>>>>> PROV-ISSUE-85 (What-is-Entity): Definition of Entity is confusing, maybe
>>>>>> over-complex [Conceptual Model]
>>>>>>
>>>>>> http://www.w3.org/2011/prov/track/issues/85
>>>>>>
>>>>>> Raised by: Graham Klyne
>>>>>> On product: Conceptual Model
>>>>>>
>>>>>> See also:
>>>>>> http://lists.w3.org/Archives/Public/public-prov-wg/2011Aug/0383.html
>>>>>>
>>>>>> Section 5.1.
>>>>>>
>>>>>> The definition of "Entity" seems to introduce un-needed complications. I
>>>>>> don't
>>>>>> see anything here that fundamentally distinguishes an entity from anything
>>>>>> that can be named, i.e. a web resource.
>>>>>>
>>>>>> I don't see what useful purpose is served by the insistence on "characterized
>>>>>> thing".
>>>>>>
>>>>>> This section seems to spend more effort describing "entity assertion" is is
>>>>>> apparently a different concept, but not formally part of the model. There is
>>>>>> some sense that an entity must have associated entity assertions... but I
>>>>>> can't see why this is needed, and indeed it may be not possible to enforce
>>>>>> this idea in RDF's open world model.
>>>>>>
>>>>>> There's been talk of Entities being part of the occurrent vs continuant
>>>>>> distinction, but I'm not seeing that explained.
>>>>>>
>>>>>> Suggest: why not just have an entity as an identifiable thing, and build the
>>>>>> rest around that? What would break with this approach?
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>
>
Received on Tuesday, 6 September 2011 11:43:23 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 13:06:41 GMT