Wikidata:Property proposal/is metaclass for

is metaclass for edit

Originally proposed at Wikidata:Property proposal/Generic

Done: is metaclass for (P8225) (Talk and documentation)

Description	all instances of the subject are subclasses of the object
Data type	Item
Domain	instance of (P31)variable-order class (Q23958852)
Allowed values	subclass of (P279)entity (Q35120)
Example 1	aircraft model (Q15056995) → aircraft (Q11436)
Example 2	cell type (Q189118) → cell (Q7868)
Example 3	ship type (Q2235308) → ship (Q11446)
Source	Wikidata:WikiProject_Ontology/Modelling
See also	metasubclass of (P2445)

Motivation edit

It seems to me like the relationship between a metaclass and the class it's about is important. Being more explicit can help us to get a more clear and consistent ontology. I prefer this way of having the relationship as it explains people who go to the item of the metaclass what it's about and it seems to me that's more necessary then explaining what the other item is about. ChristianKl ❪✉❫ 17:42, 22 March 2020 (UTC)[reply]

Discussion edit

WikiProject Ontology has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead. ChristianKl ❪✉❫ 17:53, 22 March 2020 (UTC)[reply]

Support --Tinker Bell ★ ♥ 16:50, 23 March 2020 (UTC)[reply]
Support Good idea, yes I think this has been missing. ArthurPSmith (talk) 18:20, 23 March 2020 (UTC)[reply]
Support --SilentSpike (talk) 12:47, 24 March 2020 (UTC)[reply]
~~Oppose~~ use subclass of (P279) instead. --- Jura 13:15, 27 March 2020 (UTC)[reply]
- Can you give an exam of how this could work? I don't believe it makes sense --SilentSpike (talk) 16:30, 27 March 2020 (UTC)[reply]
  - It's currently used that way on aircraft family (Q15056993). Also the ones above. --- Jura 08:34, 28 March 2020 (UTC)[reply]
- Subclass seems very wrong to me. Subclass implies that there's inheritance of the attributes and Airplane shouldn't inherit the instance of (P31) first-order metaclass. ChristianKl ❪✉❫ 10:52, 29 March 2020 (UTC)[reply]
  - @Jura1: This was what I couldn't figure out how to say above. Do you still think P279 is appropriate here? --SilentSpike (talk) 16:50, 3 April 2020 (UTC)[reply]
- subclass of (P279) is definitely wrong. If one class is subclass of (P279) of another, then instances of the first class are also instances of the second, so they are at the same level of the instance hierarchy (speaking a bit loosely). From woman (Q467) subclass of (P279) human (Q5) we get that all women are humans. But instances of cell type (Q189118) (e.g., red blood cell (Q37187)) are not instances of cell (Q7868) (e.g., an actual red blood cell in my body). Of course there are cases where this relationship is (or hopefully was) incorrectly stated in Wikidata, such as aircraft family (Q15056993). Peter F. Patel-Schneider (talk) 17:34, 3 April 2020 (UTC)[reply]
  - Changed it to Neutral --- Jura 18:34, 4 April 2020 (UTC)[reply]
@Peter F. Patel-Schneider: what do you think? --- Jura 17:09, 3 April 2020 (UTC)[reply]
Neutral There is something to be said for this proposal, but I think that there is not enough said about the property here. What makes a class a metaclass for another? Is it just the constraint stated in the proposal? If so, why not just add the constraint directly (perhaps with the addition of the ability to have named constraints)? If not, what else goes into the metaclass relationship? Adding a new property that affects the ontological backbone of Wikidata requires a full examination of how it will affect Wikidata, in my opinion. And then there should be an examination of the current content of Wikidata, or at least a good part of it, to see where this property should be added and what parts of Wikidata should be changed to align with the proposed change. Peter F. Patel-Schneider (talk) 17:43, 3 April 2020 (UTC)[reply]
Comment metasubclass of (P2445) was just added to "see also". Ironically its name includes "subclass". If it was actually used, we would probably have to change queries from P279* to (wdt:P279/wdt:P2445)* or add more P279 --- Jura 18:34, 4 April 2020 (UTC)[reply]
- Hmm. Yet another sub-property that should be set up better. It would be nice to have [...] a constraint added to [express the complex meaning of this property]. Edit: Ooops, it appears that metasubclass of (P2445) isn't supposed to mean what I thought its name implied so the required constraint is more complex than I had thought, thus the changes signalled by the []s. Peter F. Patel-Schneider (talk) 23:53, 4 April 2020 (UTC)[reply]
- @Peter F. Patel-Schneider: Constraints are aspects that can be added to properties in Wikidata but that don't exist for items. Wikidata currently doesn't use metaclass relations in a consistent fashion and there a need to invest work to get consistent definition in a lot of places. I don't think it makes sense to discuss all the case beforehand but first decide on the principles of wanting to have metaclass relationships as doing all that analysis is prohibitive for actually getting things done. ChristianKl ❪✉❫ 14:42, 12 May 2020 (UTC)[reply]
- I agree that there needs to be a discussion of principles first. There are lots of places where a discussion of principles (not necessarily agreement, just discussion) would help in Wikidata. Peter F. Patel-Schneider (talk) 15:26, 12 May 2020 (UTC)[reply]
Support I would agree more could be said about this, but also think we should not avoid it until we say more. It makes sense and would make validation easier. Iwan.Aucamp (talk) 09:55, 17 May 2020 (UTC)[reply]
Comment @ChristianKl: I added some missing parameters to the template, please raise objections if you have any. Iwan.Aucamp (talk) 22:04, 19 May 2020 (UTC)[reply]
@Iwan.Aucamp: My intuition would be that plenty of first-order metaclasses are instances of second-order metaclasses and not of variable-order class (Q23958852). ChristianKl ❪✉❫ 22:13, 19 May 2020 (UTC)[reply]
@ChristianKl: but would instance of (P31)variable-order class (Q23958852) not still hold transitively for those cases, maybe I should clarify the statement was intended to be transitive? Iwan.Aucamp (talk)
@Iwan.Aucamp:I don't consider instance of (P31) transitive. If A instance of (P31) first-order metaclass and A instance of (P31) B with B instance of (P31) second-order metaclass, A isn't a second-order metaclass. Our current way to mark transitivity of properties is to give them instance of (P31) transitive Wikidata property (Q18647515) and we don't classify instance of (P31) that way. ChristianKl ❪✉❫ 22:33, 19 May 2020 (UTC)[reply]

@ChristianKl, Tinker_Bell, ArthurPSmith: @SilentSpike, Jura1, Peter_F._Patel-Schneider: Done please have a look and raise any issues on talk page, especially with constraints which are possibly wrong. Iwan.Aucamp (talk) 22:56, 19 May 2020 (UTC)[reply]

WikiProject Ontology has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead.