Wikidata:Property proposal/disjoint with

disjoint with

   Not done

Description	disjoint with
Data type	Item
Domain	any class
Allowed values	only classes, not individuals
Example	time interval (Q186081) → relative location (Q23008351)
Source	https://www.w3.org/TR/owl-ref/#disjointWith-def

There is a property "different from" (P1889) which is defined as equivalent to owl:differentFrom. According to OWL, "different from" should only be used for individuals, not for classes. owl:disjointWith should be used to assert that no member of a class can be a member of the disjoint class. This is useful for reasoning in property path queries. Chjohnson39 (talk) 12:16, 5 May 2016 (UTC)

Discussion

•  Comment Any example where Disjoint union of Search would not fit ? author  TomT0m / talk page 12:26, 5 May 2016 (UTC)
•  Support --Emitraka (talk) 21:31, 31 May 2016 (UTC)
•  Support This is a property I might use. Definitely makes sense.--Andrawaag (talk) 21:31, 31 May 2016 (UTC)
•  Support Egon Willighagen (talk) 17:49, 2 June 2016 (UTC)
•  Comment There is extensive discussion regarding "disjoint with". --Izno (talk) 12:30, 3 June 2016 (UTC)
•  Oppose, per TomT0m. If this intends to replace disjoint union of (P2738), we need to have a discussion about that first. If not, then it would just be redundant, unless someone can think of a counter-example. --Yair rand (talk) 11:18, 6 June 2016 (UTC)
•  Oppose per Yair rand, until the question by him and TomT0m have been addressed. --Srittau (talk) 20:09, 11 June 2016 (UTC)
  •  Support per the explanation by Vladimir below. @TomT0m, Yair rand, Pasleim: Please have a Look again. --Srittau (talk) 02:09, 10 August 2016 (UTC)
•  Oppose per TomT0m --Pasleim (talk) 14:54, 24 June 2016 (UTC)
•  Support "Disjoint union of" is a modeling construct that says how a parent class is composed of its subclasses. "Disjoint with" is a constraint on two classes, that may otherwise be unrelated in the class hierarchy. It's a useful validation constraint. --Vladimir Alexiev (talk) 13:52, 11 July 2016 (UTC)
  • @Vladimir Alexiev: (and @Srittau: )This is highly mitigated by the fact that disjointness is inherited by subclassing : If A is disjoint with B, which can be established in one of the hihest level of the classification tree, then it's redundant to express that any subclass of A is disjoint with any subclass od B. For example, if we say that "event" is disjoint with "object", it's useless to repeat that "human" is disjoin with "football game". My guess would be that we can very often find some high level class (or create it if it's really needed) that make a lot of raw "disjoint with" claim totally useless. And that this would be kind of naturally sorted and would reduce the mess quite a lot, as there is a natural place to put the claim, on the last common superclass item, and not on any of the subclass combination. author  TomT0m / talk page 08:33, 11 August 2016 (UTC)

 Weak support yes, this seems quite different from disjoint union of (P2738), though in practice I'm not sure how easy it would be to use in reality given the open nature of wikidata and the messiness of the world. With the given example time interval (Q186081) expected to be disjoint with relative location (Q23008351) - that statement would be violated by creating a space-time interval that was a subclass of both, right? I'd like to see some more practical examples of where this would be helpful. ArthurPSmith (talk) 15:03, 4 August 2016 (UTC)

 Weak oppose Why would it be interesting to add that time interval (Q186081) and relative location (Q23008351) are disjunct? We have different from (P1889) for cases that might be misunderstood. So the given example is not convincing to me. Maybe, but not really, position (Q4164871) and relative location (Q23008351) could be a case, in English at least. But I think that is just simple homonymy. So, how is this property going to be relevant? Lymantria (talk) 08:29, 13 August 2016 (UTC)

@Chjohnson39:  Not done, no consensus --Srittau (talk) 00:01, 20 August 2016 (UTC)