Wikidata:Property proposal/property describes - Wikidata

Originally proposed at Wikidata:Property proposal/Generic

Originally proposed at Wikidata:Property proposal/Generic

Originally proposed at Wikidata:Property proposal/Generic

We have 12,141 properties ... we currently categorize them mainly via instance of (P31), e.g. applies to jurisdiction (P1001)instance of (P31)restrictive qualifier (Q61719275) & object named as (P1932)instance of (P31)non-restrictive qualifier (Q61719274). The problem with describing this via instance of (P31) is that it forces data consumers to resolve all subclasses of all instance of (P31) values just to see if any of the superclasses is restrictive qualifier (Q61719275). I think there's a better way to describe them via dedicated properties: I think we can attribute each function of a property within a specific property scope (as main value (Q54828448), as qualifier (Q54828449) or as reference (Q54828450)) to one specific category, namely:

• properties used as main value (Q54828448) either:
  • describe the subject in particular (such statements generally aren't inherited by instances and subclasses of the subject), or
  • describe a general characteristic of the subject (such statements likely are inherited by instances and subclasses of the subject ... note that inherited statements can be negeted with negating properties)

  (This currently isn't modeled at all, which is unfortunate, see also this discussion)

• properties used as qualifier (Q54828449) either:
  • describe the subject of the qualified statement in a non-restrictive manner, or
  • describe the object of the qualified statement in a non-restrictive manner, or
  • describe the whole qualified statement (not just the subject or object of the statement) in a non-restrictive manner, or
  • restrict the scope of the qualified statement

  (Currently modeled via restrictive qualifier (Q61719275) & non-restrictive qualifier (Q61719274) created by @Swpb: and Wikidata qualifier to describe the subject of a statement (Q115396232) & Wikidata qualifier to describe the object of a statement (Q115396176) created by me)

• properties used as reference (Q54828450) either:
  • clearly identify the referenced work, or
  • describe the referenced work, or
  • describe where the reference can be located within the referenced work, or
  • describe the quotation from the referenced work that supports the statement, or
  • describe where the statement was inferred from

  (This currently doesn't seem to be modeled at all.)

So I think it would make sense to introduce three properties "describes when used as a (main property|qualifier|reference)" these properties would all be constrained to:

• only have one value via single-value constraint (Q19474404)
• only have the values for the respective scope listed above via one-of constraint (Q21510859)

I want to thank @Lectrician1: for pointing out that our usage of instance of (P31) to categorize properties is suboptimal :)

Cheers, --Push-f (talk) 10:00, 12 December 2022 (UTC)[reply]

•   WikiProject Properties has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead. --Push-f (talk) 10:03, 12 December 2022 (UTC)[reply]
• If you propose a property that has the item datatype, please create examples that use that datatype instead of using strings.

We have properties that can be used as in different ways. The properties should likely be named like 'describes when used as qualifier' and 'describes when used as reference'. ChristianKl ❪✉❫ 13:44, 12 December 2022 (UTC)[reply]

Thanks, I have implemented these two suggestions (creating items for the suggested values and relabeling the proposed properties). --Push-f (talk) 05:26, 13 December 2022 (UTC)[reply]

• I don't think that it's a good idea to use a property along the lines of "main property describes" to describe transitivity. I also don't believe the examples you brought should be listed as being transitive over instance of (P31). The main property by it's nature always describes the subject, so I don't see a good reason for "main property describes". ChristianKl ❪✉❫ 13:44, 12 December 2022 (UTC)[reply]

  As I explained in the data model discussion this is not transitivity and cannot be modeled via value hierarchy property (P6609).

  Let me give you an example for why I think the distinction I am trying to make with "describes when used as a main property" is desperately needed.

  What's the purpose of a DeLorean DMC-12 (Q370524)? What parts does a DeLorean DMC-12 (Q370524) have? What is DeLorean DMC-12 (Q370524) named after?

  DeLorean DMC-12 (Q370524) is a subclass of sports car (Q274586), motor car (Q1420), motor vehicle (Q752870) and vehicle (Q42889).

  From DeLorean DMC-12 (Q370524)subclass of (P279)vehicle (Q42889) & vehicle (Q42889)has use (P366)transport (Q7590) a data consumer should be able to infer DeLorean DMC-12 (Q370524)has use (P366)transport (Q7590).

  From DeLorean DMC-12 (Q370524)subclass of (P279)motor car (Q1420) & motor car (Q1420)has part(s) (P527)engine (Q44167) a data consumer should be able to infer DeLorean DMC-12 (Q370524)has part(s) (P527)engine (Q44167).

  To be able to make such inferences it is crucial for a data consumer to be able to determine whether a property describes the subject in specific (e.g. described by source (P1343), named after (P138) or image (P18)) or a general characteristic of the subject that is shared by instances and subclasses of the subject (e.g. has use (P366) & has part(s) (P527)).

  For example the claims image (P18)Porsche 911 Carrera S (7522427256).jpg and named after (P138)sport (Q349) on sports car (Q274586) obviously do not apply to DeLorean DMC-12 (Q370524).

  --Push-f (talk) 10:13, 13 December 2022 (UTC)[reply]

•   Question Would it be possible to achieve the intended outcome using qualifiers on the statements about classes, for example something like vehicle (Q42889)has use (P366)transport (Q7590)nature of statement (P5102)<instance characterization> (the class vehicle (Q42889) does not has use (P366)transport (Q7590), rather its use is something like ‘grouping’ or ‘abstraction’) and sports car (Q274586)named after (P138)sport (Q349)nature of statement (P5102)<class characterization>? (Compare, in object-oriented programming (Q79872), the distinction between class variable (Q3275545) and instance variable (Q2517175).) ―BlaueBlüte (talk) 09:09, 31 January 2023 (UTC)[reply]
•   Comment As for the property proposed to be used with regard to references, I think there is the larger question of how properties ranging from those identifying works to parts of works to increasingly fine-grained locations (volume (P478), chapter (P792), section, verse, paragraph, or clause (P958), page(s) (P304), proposed ‘footnote’ (Pxxx), …) interact with one another—which ones create contexts or namespaces for which subsequent more fine-grained ones. For example, the proposal does not seem to address how one would indicate that a page number is relative to the whole work and not to the volume, or the issue and not the volume, or whether a footnote number is specific to the chapter or just happens to appear in that chapter. (See also possible discussion regarding this problem with respect to footnotes here.) The Source MetaData WikiProject does not exist. Please correct the name. ―BlaueBlüte (talk) 21:01, 31 January 2023 (UTC)[reply]

  The Source MetaData/More WikiProject does not exist. Please correct the name. ―BlaueBlüte (talk) 21:01, 31 January 2023 (UTC)[reply]

•   Comment @Push-f: Can you motivate better why we need new props for this, rather than just new items? It seems the only motivation you give is purely technical:

"resolve all subclasses of all instance of values just to see if any of the superclasses is restrictive qualifier"

Are you concerned by performance, or the difficulty of writing a property path in SPARQL? --Vladimir Alexiev (talk) 19:34, 13 February 2023 (UTC)[reply]

  Not done unanswered questions for months @Vladimir Alexiev, BlaueBlüte, Push-f, ChristianKl, Lectrician1: BrokenSegue (talk) 06:04, 29 June 2023 (UTC)[reply]