Wikidata:Property proposal/reference has role

reference has role

Originally proposed at Wikidata:Property proposal/Generic

   Done: reference has role (P6184) (Talk and documentation)

Description	role, or specific nature, of the given reference
Represents	reference (Q121769)
Data type	Item
Domain	references
Allowed values	limited list of allowed values
Example	AS PART OF REFERENCE → first valid description (Q1361864)
Robot and gadget jobs	convert existing uses of P31 as a reference to use the new property
See also	subject has role (P2868), object has role (P3831), type of reference (P3865)

Motivation

We currently have about 50,000 cases where instance of (P31) is used as a property in a reference, rather than as a main statement. See queries: tinyurl.com/y78odkdu (counts of values), tinyurl.com/y8owabzc (examples). The use on e.g. Q101538#P225 is typical.

In my opinion, use of P31 in this way is ugly and confusing -- IMO it would be better if P31 was only used for its main purpose, as a direct statement on an item giving its nature.

Use of P31-on-references is similar to the way P31 once used also to be used as a qualifier. But those uses have now everywhere been removed and replaced with subject has role (P2868) and object has role (P3831).

P31-on-references is currently doing some important work. In particular, the #1 value of P31-on-references, first valid description (Q1361864), to be able to indicate that the reference in question contained the first description and definition of a taxon is extremely valuable and important to be able to highlight for taxonomic references.

To me it therefore makes sense to propose a new drop-in replacement "reference has role" for P31-on-references, as a specific property to take over this function, which is different from the normal use of P31; and which would allow a constraint to limit acceptable values to an agreed controlled vocabulary. -- Jheald (talk) 14:30, 10 June 2018 (UTC)

Discussion

• Proposed. Jheald (talk) 14:30, 10 June 2018 (UTC)

WikiProject Taxonomy has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead. -- Jheald (talk) 14:56, 10 June 2018 (UTC)

•  Support. Appears to me, a relatively inexperienced Wikidata-er, to fill a niche role. If/when created, could instance of (P31) be software-restricted from being placed in references, to avoid confusion & accidental placement? —Tom.Reding (talk) 15:24, 10 June 2018 (UTC)

The software wouldn't completely prevent it, but it would register a constraint violation, and place a warning error sign next to it. Jheald (talk) 19:52, 10 June 2018 (UTC)

 Comment - This is not an isolated proposal: publication in which this scientific name was established also deals with this issue, but proposes to move this to a statement. Since there are so many cases for this, and likely to become more, a separate property (making this a statement) seems well justified. The qualifier "first description (of a taxon) (Q1361864)," looks quite awkward to me (also wrong: it is the establishing of a name that matters here, not the description): if there are three references listed, will the software that reads in data be able to determine what reference this qualifier belongs to? A separate property would make this unnecessary. - Brya (talk) 17:12, 10 June 2018 (UTC)

@Brya: I don't see the two proposals as necessarily in conflict. Firstly, there are other reasons why one might want to annotate a reference: establishing a taxon is only one example. Secondly, even if there was a separate statement for when the scientific statement was established, one might still want to note of a reference that it was the originating paper, or the statement that redefined the taxon, or some other notable thing about the paper. But it would be a good thing to get rid of the current P31s.

As to your technical question, the annotation becomes part of the reference (as the present uses of P31 do). It is therefore uniquely associated with a single reference, just as much as the properties "stated in" or "volume" or "page" would be. There is no danger of crosstalk with any other reference. Jheald (talk) 19:52, 10 June 2018 (UTC)

Yes, on this last point, I had realized that the danger of software reading it out wrong was quite limited. It will still be confusing to the reader. - Brya (talk) 16:41, 11 June 2018 (UTC)

Hi Brya, I think this proposal makes queries like that a little bit more understandable. I'm not really happy with the creation of P5326 (P5326). --Succu (talk) 19:10, 25 June 2018 (UTC)

Hi Succu, I understand that switching to P5326 (P5326) would involve a lot of edits (some fifty thousand items being involved). But there is an even larger number of items involved which don't yet have an original publication attached (much, much larger), so it is worth taking time to reconsider before taking on that larger number. Having a property like P5326 is much more user-friendly. - Brya (talk) 05:31, 26 June 2018 (UTC)

•  Support. This is a needed addition to avoid ambiguity. It is a logical companion to object has role (P3831) and nature of statement (P5102). Deryck Chan (talk) 21:27, 10 June 2018 (UTC)
  • Moving to  Neutral for now because of the competition with type of reference (P3865) below. @Succu, Jheald: Will you please explain where you draw the boundary between this new property and type of reference (P3865)? Which ones of the top 10 hits of tinyurl.com/y78odkdu do you intend to use with this new property? Deryck Chan (talk) 12:05, 2 July 2018 (UTC)

This add (made by Jheald) would be a perfect usage of type of reference (P3865). For my concerns please see below (type <> role). --Succu (talk) 21:45, 4 July 2018 (UTC)

•  Comment type of reference (P3865) could work for this.
  --- Jura 17:20, 15 June 2018 (UTC)

•  Support I'm thinking about this for a while. --Succu (talk) 12:39, 16 June 2018 (UTC)
•  Oppose per above.
  --- Jura 08:57, 17 June 2018 (UTC)
  • There is a difference between type and role. For type of reference (P3865) = scholarly article (Q13442814) the role could be emendation (Q1335348). --Succu (talk) 10:12, 18 June 2018 (UTC)
    • I don't think there is a need for both. Type can be first valid description (Q1361864).
      --- Jura 06:27, 19 June 2018 (UTC)
      • first valid description (Q1361864) is not necessarily a subclass of scholarly article (Q13442814)… --Succu (talk) 19:46, 19 June 2018 (UTC)
    • If that difference is needed, a specific item about the reference should be made instead. In general, I don't think that matters.
      --- Jura 08:35, 5 July 2018 (UTC)
      • Sorry Jura, but I don't understand your remark. What „specific item about the reference should be made“ in the case of Linnés book? --Succu (talk) 20:26, 5 July 2018 (UTC)
        • The work already has an item, so why would you want to use P3865 to described that it's nature? The question might arise if only "reference url", not "stated in" is present.
          --- Jura 04:16, 6 July 2018 (UTC)
•  Support In the sense that Succu proposed. I don't think it makes sense to automatically map the instance of (P31) values to the new property as type of reference (P3865) should often be nearer to that.

first valid description (Q1361864) seems to me qualitatively different from our existing usage of type of reference (P3865). ChristianKl ❪✉❫ 21:04, 7 November 2018 (UTC)

@Deryck Chan, Tom.Reding, ChristianKl, Brya, Succu, Jheald: @Jura1:  Done: reference has role (P6184). − Pintoch (talk) 06:36, 27 November 2018 (UTC)