Hi Thomas,

Yes, your example is the kind of use case.  And disagreements often come up with the  "then C is a part of B".

And I've found that many times the classification problems where "wait a second, this is not actually quite true, and the right class in the hierarchy I'm thinking of" ... will typically be resolved, and consensus more easily agreed by all parties when the class is either:
A. made more abstract for the benefit of all, but saving that...
B. make a new class, more broader, and less disagreeable by all parties

It's just a simple data modeling problem oftentimes, but where "transitive over" seems to expose them.
I see it more about getting consensus and the "it depends" seemed oftentimes easily solved with a different broader class used or created and applied.



On Sun, May 22, 2022 at 8:25 AM Thomas Douillard <thomas.douillard@gmail.com> wrote:
Sounds not uninteresting, but without the context of the discussion it’s hard to understand what the problem actually is. Could include more details ?

I googled a bit and found that a « transitive over » usecase could be
If
* researcher Livingstone « explores » Egypt
* Egypt is a part of Africa
then we can conclude that researcher livingstone explores Africa

This would mean that the « explore » property would be transitive over « part of ».
An example of this could be of course « instance of » who is of course transitive over « subclass of ».

Yet « subclass of » is of course a transitive property by itself. If all mammals are animals (so ''mammal subclass of animal')' and all animals are organism ''animal subclass of living organism'', then of course all mammals are living organisms.

As for transitivity of « part of » of course if A is a part of B and C is a part of A, then C is a part of B. The thing is that someone that studies, say a lineage of cell of some kind of animal could not necessarily considering to study the animal itself, so « studies » could not be considered to be transitive over « part of ».

What exact reasoning problem do you have in mind ?

Le sam. 21 mai 2022 à 17:56, Thad Guidry <thadguidry@gmail.com> a écrit :
I wanted to share my reply to a recent Telegraph conversation:

Thad Guidry, [5/21/2022 10:22 AM]
[In reply to Nikki]
Agree somewhat, however in the case of P31 we already have P6609 that describes the general SKOS/OWL "transitive over" and we added the value-type constraint https://www.wikidata.org/wiki/Q21510865 to be transitive property https://www.wikidata.org/wiki/Q18647515

But that was not the case with P279 ... where instead we stated that P279 itself is an instance of transitive property ... which is what probably confuses folks.

[[wikilinksbot]], [5/21/2022 10:22 AM]
P31 (https://www.wikidata.org/entity/P31) – instance of
P6609 (https://www.wikidata.org/entity/P6609) – value hierarchy property
P279 (https://www.wikidata.org/entity/P279) – subclass of

Thad Guidry, [5/21/2022 10:26 AM]
So P279 is a https://www.wikidata.org/wiki/Q18647515 and P31 is not.

[[wikilinksbot]], [5/21/2022 10:26 AM]
P279 (https://www.wikidata.org/entity/P279) – subclass of
P31 (https://www.wikidata.org/entity/P31) – instance of

Thad Guidry, [5/21/2022 10:29 AM]
Details here:  https://www.w3.org/TR/owl-ref/#TransitiveProperty-def

Thad Guidry, [5/21/2022 10:38 AM]
So... (lolol) .... through transitivity once an item becomes an instance of a class... then it automatically inherits all properties of that class... but only and strong ONLY WHEN it is considered an instance of a class... and not before.

Reasoners, interpreters (external, custom code, institutions, etc.) might apply transitivity "slightly" differently for different contexts, and might bucket some items prematurely to be considered an instance of a class ... but generally, the old adage is that of the above paragraph... only once it is considered an instance of.

The problem as often seen in Wikidata is that sometimes higher classes are currently not abstract enough sometimes to fulfill broader roles... and hence... a broader higher class oftentimes just needs to be created to make things in the hierarchy a bit more sensical.

_______________________________________________
Wikidata mailing list -- wikidata@lists.wikimedia.org
To unsubscribe send an email to wikidata-leave@lists.wikimedia.org
_______________________________________________
Wikidata mailing list -- wikidata@lists.wikimedia.org
To unsubscribe send an email to wikidata-leave@lists.wikimedia.org