Re: Different senses of specialization

Hi James
Your definitions are good in the context of the semantics where Thing (capital T) is defined.

It does not immediately translate to a version with 'thing', for the DM,
where entity is defined as:

"An entity is a physical, digital, conceptual, or other kind of thing; entities may be real or imaginary."

Essentially, we would have: entity is a thing   vs   entity is an aspect of a thing.


Professor Luc Moreau
Electronics and Computer Science
University of Southampton
Southampton SO17 1BJ
United Kingdom

On 11 Apr 2012, at 12:04, "James Cheney" <jcheney@inf.ed.ac.uk<mailto:jcheney@inf.ed.ac.uk>> wrote:

Hi,

Specialization and alternate are relations between entities in the semantics.

Saying that "entities are things" is vague: does it mean "all entities are things", or does it also mean "entities and things are the same concept"?  Can you please refer me to the resolution text on this point?

Does this mean that "all entities are things"? This is sensible, and is allowed but not required by the semantics.

Conversely, does it mean "all things are entities"?  We've been using "Thing" to talk about things that can change over time, whereas entities are fixed aspects of such things - which we need in order to provide provenance records that are stable over time.  So if we say "all entities are things, AND all things are entities" then this distinction disappears and becomes purposeless.

An Entity could be a Thing in the semantics.  The semantics doesn't say whether entities are things or not.  All it prescribes are

- each Thing has a lifetime and some attributes that can vary
- each Entity has a lifetime and some attributes which don't vary
- each Entity is formally linked to a Thing with a larger lifetime, by a function thingOf.

: the sets of entities and things could be any sets and could overlap.  However, any Thing in an instance of the semantics that chances over time cannot be an Entity.

One can think of the thingOf function in different ways.  I tend to think of the entities as "information about things", with the thingOf function linking the entity to the thing that it is about, but I understand that this perspective isn't widely shared.  I'm happy to rephrase it if we can find a more agreeable term.

Would these rephrasings be OK with you (and others, especially Tim):

"alternateOf: To express when one Entity is an aspect of the same Thing as another Entity. "

"specializationOf: To express when one of two alternate Entities is more a specific aspect of the Thing they are both based on"

--James

On Apr 11, 2012, at 6:14 AM, Luc Moreau wrote:

Hi James,

I have some difficulty with your descriptions of specialization.

"specializationOf_1: To express when one entity provides a more specific description of the same thing as another entity. "

This seems to indicate that an entity is a description of a thing.

It thought that an entity *IS* a thing, and we provide descriptions for entities. So, I don't think that an entity
describes something.

A long time ago, I asked if specialization/alternate were relations between descriptions or between entities. The
group responded these relations were between entities!

Given this, does your classification still hold? Can it be phrased differently?

Luc

On 11/04/2012 05:43, Luc Moreau wrote:
Tracker, this is ISSUE-29.

Professor Luc Moreau
Electronics and Computer Science
University of Southampton
Southampton SO17 1BJ
United Kingdom

On 9 Apr 2012, at 15:58, "James Cheney"<jcheney@inf.ed.ac.uk<mailto:jcheney@inf.ed.ac.uk>>  wrote:


Hi,

In reading the discussion over the last week and while reviewing the various documents, I've noticed that "specialization" seems to be being used in (at least) three different ways.  I think this is a contributor to the confusion concerning specialization and alternate.


specializationOf_1: To express when one entity provides a more specific description of the same thing as another entity.

Example: The entity ex:article in the primer, with its versions ex:articleV1 and ex:articleV2, both of which are specializations of ex:article.

This is the sense that I have been assuming and it is what is reflected in the semantics.


specializationOf_2: To express when one entity provides more specific information than another (which need not be about the same thing!)

Example: In PROV-DM-CONSTRAINTS:

specializationOf(customerInChairAt6pm, customerInChair)
specializationOf(customerInChairAt7pm, customerInChair)

To me this does not make sense because the entities customerInChairAt6pm and customerInChairAt7pm are about different "Things" (the different customers Alice and Bob).  So there is no single Thing to which customerInChair describes, contradicting (what I thought was) the consensus that an entity describes an aspect of exactly one thing.

In my view, this sense of specialization is specious, since one entity may provide "more specific" information than another entity but the two entities need not be about the same thing, as the relationship may be purely coincidental.  To carry this to an extreme, any entity is a "specializationOf(2)


specializationOf_3: (work / item) To express when one entity describes a more concrete thing that is an instance of a more abstract thing described by another entity.  This is like the work/item distinction in FRBR.

Example: The primer refers to a file (which can have multiple versions), and a specific copy of a file on  a hard disk, as another example of specialization.

To me this does not make sense because a file, in the abstract, is just a sequence of bits, which could be physically realized anywhere or could be realized in multiple places at once.  A file-on-disk might carry the same information as an abstract-file, but can only exist in one place.  We do not talk about locations explicitly in PROV, but even so, I think this is confusing: again, to carry the argument to an extreme, the number 2, in the abstract, is not the same kind of thing as a piece of paper with "two" written on it.

I think specializationOf_3 ("instance/realization of") is a sensible notion, but it should not be conflated with specializationOf_1.  There are already vocabularies that deal with this type of relationship, such as FRBR itself.  If we believe that PROV should attempt to solve this in a new way, I think we should avoid overloading the notion of specialization as "different, more specific information about the same thing" with this notion, which is really about "more concrete instance of an abstract thing"



To summarize:
- I think we should be careful about these different senses of "specializationOf".
- Only the first sense is supported by the current version of the semantics.
- If we can agree on one of these definitions for specializationOf, but believe other senses need to be modeled, we should introduce additional relations to name them, and ensure that the meanings are clear and they are used consistently in examples.


--James
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.







--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.