how to restrict a superproperty to a subproperty

Martin Doerr (pc) 7 Feb 2024 

(...) 

I also encountered that we can solve an old problem elegantly, the "restricted IsA":

Frequently, we specialize a class and its property, meaning that this superproperty may not explicitly be instantiated for the subclass. The FOL is easy: Just write that subclass and superproperty implies subproperty.

Best,

Martin

Christian-Emil Ore (pc) 15 February 2025

The issue 655 is

In the 57th CIDOC CRM & 50th FRBR/LRMoo SIG Meeting, the SIG resolved to start an issue where to determine if any unidirectional implication expressed in some property’s FOL can be called a “shortcut”, irrespective of its complexity, or if the term can only be applied to straightforward paths. To be discussed in a new issue and to be brought at the next SIG meeting for discussion.

 

It should be sufficient to use the P7 implication as shown in my diagrams to state that "the term [shortcut] can only be applied to straightforward paths.

Here you need to be more explicit:
" I also encountered that we can solve an old problem elegantly, the "restricted IsA":

Frequently, we specialize a class and its property, meaning that this superproperty may not explicitly be instantiated for the subclass. The FOL is easy: Just write that subclass and superproperty implies subproperty.

This is an old question: For classes A, B and A' where A' is a subclass of A, P, A -> B, P': A' -> B where P' is a subproperty of P is P identical to P' when P us restricted to P'? It is clear that P'(x,y) implies P(x,y)  from the rules for FOL formulations. The question is if P(x,y) and A'(x) implies P'(x,y) or not.  My impression is that this issue has been discussion once a while in the last 20 years but no explicit decision has been taken. The cleanest and most common practice in object oriented modelling and programming is that the superproperty is identical to the subproperty when restricted to the domain and range of the subproperty.  I suggest we introduce this formally in CRM.

Best,
Christian-Emil 
 

Christian-Emil Ore (pc) 1 March 2024

The discussion was a little diverted to a discussion about strict inheritance. This may be another issue, but I suggest we add under the entry for subproperty in the terminology section (between point 5 and 6 in the list):

the superproperty is identical to the subproperty when restricted to the domain and range of the subproperty

The will require the following adjustment of the FOL exemplified by  P32 below.

E7(x) ∧ E55(y) ∧ P2(x,y) ⇒ P32(x,y)

or better (?)

E7(x) ∧ E55(y)  ⇒  [P2(x,y) ⇔ P32(x,y)]

Wolfgang, opinions?

Shortcuts and complex connections.

I have checked again my simple diagrams in https://cidoc-crm.org/sites/default/files/P156_P7_inverse_shortcut_long%20paths.pptx 
for 
P7(x,y) ⇒ (∃z,u) [E53(z) ∧ E18(u) ∧ P157i(y,z)) ∧ P157i(u,z) ∧ P161(x,z) ∧ P89(z,y) ]
In the scopenote:
Therefore, this property implies the more fully developed path from E4 Period through P161 has spatial projection, E53 Place, P89 falls within to E53 Place, where the intermediate place is also defined in the same geometric system.

P156 occupies (is occupied by)
P156(x,y) ⇔ (∃z) [E18(x) ∧ E53(y) ∧ P196(x,z) ∧ P161(z,y) ∧ P157(y,x)]
In the scopenote:
Therefore, this property implies the more fully developed path from E4 Period through P161 has spatial projection, E53 Place, P89 falls within to E53 Place, where the intermediate place is also defined in the same geometric system.

They are not considered to be shortcuts, just implications. They could perhaps be called conditional shortcuts with restrictions on the shortcut property while ordinary shortcuts should be unconditional. This may have consequences for other shortcuts?

Best,
Christian-Emil

Martin Doerr (pc) 1 March 2024

Dear Christian-Emil,

A) Yes, the issue of restricting superproperties to a subproperty is a different issue. 
I will present a series of such statements for harmonizing reification with Named Graphs of factual statements(!).

I agree with the dformulation of the constraint:
E7(x) ∧ E55(y)  ⇒  [P2(x,y) ⇔ P32(x,y)],

but not that any type of activity is a general technique.

Obviously, restricting a superproperty implies that the superproperty is not necessary. I further maintain  that restricting a superproperty to a subproperty is case-specific semantics.

See also: https://www.sciencedirect.com/science/article/abs/pii/S0306437998000015 

(https://publications.ics.forth.gr/tech-reports/1997/1997.TR198.Inheritance_Particip_Constraints_Disjointness.ps.gz)

B) We should not argue with the term "implies" in the scope note:
"Therefore, this property implies the more fully developed path from E4 Period through P161 has spatial projection, E53 Place, P89 falls within to E53 Place, where the intermediate place is also defined in the same geometric system".

This formulation is nothing else than the circumscription of an "inverse shortcut" (and "strong shortcut") before we introduced the term. First we should clean up the scope notes.

I believe that two-property paths are too restrictive for shortcut semantics. We may encounter shortcuts including shortcuts, as I will show in the meeting. If we get rid of the intermediate, we have a 3 property path etc.

However, It may be advisable to use only the term "implies" for longer paths, not to confuse users with complicated definitions of what a shortcut is.

What do you think?

All the best,

Martin

Christian-Emil Ore (pc), 2 March 2024

Dear  Martin,

Firstly, "Obviously, restricting a superproperty implies that the superproperty is not necessary" The other way round?

I am happy with the fact that  a subproperty is not identical to its superproperty  restricted to the domain and range of the subproperty in our model.  Participate does not imply perform but the other way round is as it is decribed in CRM.

Ad P7:  According to my memory we decided to use "implies" and not a variant term of shortcut because the conclusion of the implication is a complex graph and not just a path.

Best,
Christian-Emil