Issue 534: Representing .1 properties of full paths in shortcut properties.

ID: 
534
Starting Date: 
2021-04-12
Working Group: 
3
Status: 
Open
Background: 

In the 49th CIDOC CRM and 42nd FRBR CRM sig meeting (virtual), upon discussing Pxxx represents instance of type the sig decided to start an issue where to discuss  how to translate  .1 properties in long paths into shorctut properties. 

Motivation:

Pxx represents entity of type is a shortcut for E36 –P138: E1 –P2: E55 AND a subproperty of P138, so it can use the .1 mode of representation
P62 depicts is a shortcut for E24-P65: E36 –P138: E1 but NOT a subproperty thereof

 

March 2021

Post by Rob Sanderson (8 June 2021)

#534 - In that RDF doesn't allow for meta-properties, we would need to reify anyway. At which point we couldn't use a shortcut. So while I understand the challenge, I don't have (or need) a solution. Seems like it would need to be on a case by case basis to determine the relationship of the meta property to the main property anyway?

In the 53rd CIDOC CRM & 46th FRBRoo SIG meeting, the Sig discussed whether it is the case that for each long path involving a .1 property, the corresponding shortcut should also come with a .1 property. 

It was established that the way to go about it is either: 

  1. to recommend that implementers only use the full path if they want to deploy the .1 property, or
  2. or, for each long path –shortcut pair, in which the long path involves a .1 property, establish whether it is needed for the shortcut too. If yes, what does it amount to? 

Decision

  • Check the pairs of long paths and corresponding shortcut properties, to determine which ones make use of a .1 property. 
    • See if there is a list of full vs shortcuts available –check issue 357 for CEOs HW & feedback by MF (posts from 21 October 2021 and on). 
    • also discuss the hierarchy of properties that have .1 properties and whether these are inherited by their subproperties (and what the relation btw the respective .1 properties) 
    • HW: CEO 

May 2022

Current Proposal: 

Post by Christian-Emil (10 September 2022)

Dear all,

Please, find my  HW attached and in the issue folder

 

 

Best,

Christian-Emil

In the 54th CIDOC CRM & 47th FRBR/LRMoo SIG Meeting, Christian-Emil Ore and presented his HW to the SIG. An outline of his proposal can be found below. 

The premise he worked on is that:

  • subproperties inherit the .1-properties from their superproperties. 
  • .1 properties do not stand in an IsA relation with the .1 version of the properties that are superproperties of their .1-less version by default. Instead decide on a case-by-case basis. 

Proposal:

  • The relation of a short path with a .1 property to a long path containing a .1 property is not predetermined (compare (a) P62.1 and P138.1 to (b) P107.1 and P144.1, respectively). 
  • If the corresponding .1 properties’ semantics appear unrelated, then restrictions need to be added to the shortcut property definition (a). Namely, if the long path has instantiated a .1 property, then the short path shouldn’t. The type on property in the long path implies any type, not the type of the property in the short path. 
  • If the corresponding .1 properties’ semantics are unrelated, then an extra FOL axiom has to be added (b).

Nb. Christian-Emil's HW can be found here

Decision: Postpone reaching a decision. In the meantime, HW to CEO (and WS because he knows his FOLs) to check data instantiating .1 properties both in the long and the short path, in order to verify that the type restrictions are as proposed. 

 

Rome, September 2022
 

Post by Christian-Emil Ore (29 November 2022)

Dear all,

My HW can be found here:

https://docs.google.com/document/d/1kd-zCO8e00WoKWb-9VAm96ECDEA2O_1iFI6k...

Best,

Christian-Emil
 

Post by Martin Doerr (30 November 2022)

Dear Christian-Emil,

I agree with your precise analysis, except for:

A)  I'd argue that P138.1 mode of representation is indeed identical to P62.1. Simply, similar examples have not been provided.  I suggest a FOL for that. 

"P138.1 mode of representation   seems to be unrelated to P62.1 mode of depiction. 

Conclusion
Neither of the .1-properties in the long path are semantically related to the .1-property of the shortcut property. Since P62 is declared as a shortcut of the long path, it cannot have the .1-property."
I agree that P67.1 is unrelated. The kind of reference of a P138 instance would be "represents". This is implicit I think in the explanation of .1 properties of "has type"-type in the introduction.  Could be a new issue. 

B) "The implication from the long path to the shortcut is ok since the shortcut has no .1-property. The implication from the shortcut to the long path is problematic since there is a .1-property in the long path. It is possible to imagine that the shortcut could have a P125.1 mode of use:E55 Type, but the range instance of P16.1 and P125 must be identical so the following additional FOL axiom would have to be added: 

P125(x,y,w) ⇔ (∃z) [E70(z) ∧ P16(x,z,w) ∧  P2(z,y)]" 

I have not understood why "the range instance of P16.1 and P125 must be identical ".

E.g., used "passiflora racemosa flower", mode of use "as model", (Martin Johnson Heade, "Hummingbirds and Passionflowers", ca 1870-1883, Museum of Fine Arts Boston, 
https://artsandculture.google.com/asset/hummingbird-and-passionflowers/L...)
a typical case for the ten thousands of botanical images from European botanical researchers in the previous centuries.

Funny example is John James Audubon: He reportedly shot a bird in 1812 in Pennsylvania, named it "Regulus cuvieri", painted it probably with an error in "Birds of America", so that this species was never seen again.

Best,

Martin

Post by Christian-Emil Ore (30 November 2022)

B:
Typo: P125  should be P125.1 in the bold phrase.

"The implication from the long path to the shortcut is ok since the shortcut has no .1-property. The implication from the shortcut to the long path is problematic since there is a .1-property in the long path. It is possible to imagine that the shortcut could have a P125.1 mode of use:E55 Type, but the range instance of P16.1 and P125.1 must be identical so the following additional FOL axiom would have to be added:

P125(x,y,w) ⇔ (∃z) [E70(z) ∧ P16(x,z,w) ∧  P2(z,y)]"

Post by Martin Doerr (30 November 2022)

OK, 

If I understand correctly, the "w" should just be propagated in both directions. If P125, nothing tells us that the unknown z is the same as one possibly in a full path of the same activity. Nothing tells us why there should not be more than one "w" for the same z. Is that what you mean?

I'd support creating such a P125.1.

Best,
Martin

Post by Christian-Emil Ore (1 December 2022)

Thank you for your comment.  The second argument of the P125 is the type of things used in an activity. Still the type of an object is not equal to the way such an object is used in an activity. Trilobite fossils can be used in the production of a jewellery as a mold. (My daughter has one in gilded silver).

This is an argument for a P125.1 mode of use.

The technicalities:
P16(x,z,w) is the current notation for P16.1(x,z,w) 

(1) for all w P16(x,z,w)  ⇒ P16(x,z)

and we already have declared the axiom
(2) P125(x,y) ⇔ (∃z) [E70(z) ∧ P16(x,z) ∧  P2(z,y)]

and (3) below can  be deduced from (1) and (2):
(3) P125(x,y) ⇐ (∃z) [E70(z) ∧ P16(x,z,w) ∧  P2(z,y)]

The other way is not so easy since (5) below cannot be deduced from (2) since the fact that a specific thing is used in an activity does not imply that this thing is used in the mode of used described by any instance of E55 Type which would have been the case if (4) were universally true.
(4)P16(x,z)  ⇒ P16(x,z,y)   
(5) P125(x,y)  ⇒ (∃z) [E70(z) ∧ P16(x,z,y) ∧  P2(z,y)]

So we need to add (6) 
P125(x,y,w)  ⇒ (∃z) [E70(z) ∧ P16(x,z,w) ∧  P2(z,y)]

That is, if an activity x, uses things of type y in the mode of w then there exists a thing used in x in the mode of w and the type of x is y. But now we need to take into consideration that the short path has a .1-property and we add 
P125(x,y,w) ⇐ (∃z) [E70(z) ∧ P16(x,z,w) ∧  P2(z,y)]

I hope this is correct. 
Christian-Emil

Post by Martin Doerr (1 December 2022)

I agree with:
 
So we need to add (6) 
P125(x,y,w)  ⇒ (∃z) [E70(z) ∧ P16(x,z,w) ∧  P2(z,y)]

That is, if an activity x, uses things of type y in the mode of w then there exists a thing used in x in the mode of w and the type of x is y. But now we need to take into consideration that the short path has a .1-property and we add 
P125(x,y,w) ⇐ (∃z) [E70(z) ∧ P16(x,z,w) ∧  P2(z,y)]
 

Post by Wolfgang Schmidle (6 December 2022)

I agree with  P125(x,y,w) ⇔ (∃z) [E70(z) ∧ P16(x,z,w) ∧ P2(z,y)] and the argument that the "w" should just be propagated in both directions.

In fact, I was wondering if the argument is specific for P125/P16 or if it can be applied to all "of type" properties that are strong shortcuts? For example P199 "represents instance of type":

short: E36 Visual Item P199 represents instance of type E55 Type
long: E36 Visual Item P138 represents [P138.1 mode of representation] E1 CRM Entity P2 has type E55 Type
P199(x,y,w) ⇔ (∃z) [E1(z) ∧ P138(x,z,w) ∧ P2(z,y)]

 

A)  I'd argue that P138.1 mode of representation is indeed identical to P62.1. Simply, similar examples have not been provided.  I suggest a FOL for that.

What does it mean that the .1 properties in a weak shortcut are identical?

short: E24 Physical Human-Made Thing P62 depicts [P62.1 mode of depiction] E1 CRM Entity
long: E24 Physical Human-Made Thing P65 shows visual item E36 Visual Item P138 represents E1 [P138.1 mode of representation] CRM Entity
P62(x,y) ⇐ (∃z) [E36(z) ∧ P65(x,z) ∧ P138(z,y)]

Do you have something like this in mind?
P62(x,y,w) ⇐ (∃z) [E36(z) ∧ P65(x,z) ∧ P138(z,y,w)]
P62(x,y,w) ∧ E36(z) ∧ P65(x,z) ∧ P138(z,y)  ⇒  P138(z,y,w)

Or in one line:
E62(x,y) ∧ E36(z) ∧ P65(x,z) ∧ P138(z,y)  ⇒  [P62(x,y,w) ⇔ P138(z,y,w)]

Wolfgang

In the 56th joint meeting of the CIDOC CRM and ISO/TC46/SC4/WG9 &49th FRBR/LRMoo SIG, it was decided that the issue be postponed to give WS & CEO more time to work on it. 

 

Crete, May 2023

In the 57th CIDOC CRM & 50th FRBR/LRMoo SIG Meeting, CEO gave a summary of where things stand wrt representing .1 properties of full paths inshortcut properties. 

For shortcuts

  • If there is a shortcut property (without a .1 property in it), whose fully developed path comes with a .1 property somewhere within it, then there is no problem 
  • If there is a shortcut property (with a .1 property in it), whose fully developed path also comes with a .1 property somewhere within it, then the implication is weakened because of the extra condition in the conclusion of the implication. The fully developed path may be true (the addition of the typed property makes no difference there), but the shortcut not necessarily so. 
  • The solution would be to check whether the .1 property can be added to the left-hand side 

For inverse shortcuts: 

  • If the .1 property only appears on the shortcut property and not on the fully developed path, then there is no problem. 
  • If a .1 property appears in the fully developed path, the dependencies between the .1 properties in the shortcut and the fully developed path should be checked (see issue 633 as well)

Discussion points: 

  • Consider every property appearing in a fully developed path that gets shortcut by another property for reflexivity, transitivity, symmetry in order to not permit vacuously creating infinite shortcut properties (as part of an unintended modeling set of constructs). 
    • F.i. if one has a gazetteer with multiple P89’s, they wouldn’t want to write down everything that is implied by them. They can be interpreted, but still wouldn’t do them. The answer would be similar to that. 
  • List of shortcuts – fully developed paths with .1 properties –has been put together by CEO for the meeting in Luxembourg. 

It is an ongoing discussion between CEO, WS & MD. To be concluded by the next meeting in Paris.
 

Marseille, October 2023

Reference to Issues: