Wikidata:Property proposal/number of persons depicted
Jump to navigation
Jump to search
number of persons depicted
[edit]Originally proposed at Wikidata:Property proposal/Creative work
Not done
| Description | persons depicted in an image, such as a painting. Can be used to check completeness of values in "depicts" (P180) |
|---|---|
| Data type | Quantity |
| Domain | items for works such as paintings, photographs |
| Example 1 | Mona Lisa (Q12418) → 1 |
| Example 2 | The Balcony (Q775407) → 4 |
| Example 3 | The Last Supper (Q128910) → 13 |
| Example 4 | The Bull (Q2917717) → 1 |
| Example 5 | Portrait of a Woman with a Squirrel (Q17335769) → 1 |
| Example 6 | The Last Supper (Q128910) → 8 based on heuristic (P887) TensorFlow (Q21447895) (fictitious sample) |
| Expected completeness | eventually complete (Q21873974) |
| Robot and gadget jobs |
|
| See also | depicts (P180), relative position within image (P2677) (most samples above are also property samples for these) |
Motivation
Based on Wikidata users' constructive input, above an improved proposal. It uses quantity datatype instead of item datatype. Apparently this is preferred. Attempts to count values of P180 directly in a reliable way weren't successful or couldn't be provided by SPARQL specialists and discussion participants. Obviously, neither was it possible to use that to check completeness of P180 itself. A sample query as a proof for the opposite would still be welcome. Values of P180 are probably more complicated to interpret than initially thought.
--- Jura 10:02, 27 July 2018 (UTC)
Discussion
Support David (talk) 07:35, 28 July 2018 (UTC)
Oppose Use "depicts", with a "quantity" qualifier; per Wikidata:Property proposal/number of people depicted. Also @Cwf97, Yair rand, Pintoch, Thierry Caro, Relf PP: from the previous discussion (which should not have been withdrawn in favour of this new proposal, since all that has changed is the datatype). Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 19:47, 28 July 2018 (UTC)
- It would be constructive input if you attempted to address the issues raised with your suggestion(s).
--- Jura 09:01, 29 July 2018 (UTC)
- It would be constructive input if you attempted to address the issues raised with your suggestion(s).
- Oppose per Andy. If you don't like depicts (P180)human (Q5), why don't you propose a property "depicts entities of the class" so that we can do depicts entities of the classhuman (Q5) instead? − Pintoch (talk) 08:52, 29 July 2018 (UTC)
- We still lack a working sample of a query. But maybe I'm just too much focused on trying to find a working solution. It would help to see a query that works for at least three images and doesn't imply redoing all statements. BTW a person isn't necessarily a human.
As for the alternate proposal, I think it would duplicate P180.
--- Jura 08:59, 29 July 2018 (UTC)
I looks like Pintoch doesn't want to their arguments to be discussed. I will attempt to ignore it going forward.
--- Jura 09:42, 29 July 2018 (UTC)
- We still lack a working sample of a query. But maybe I'm just too much focused on trying to find a working solution. It would help to see a query that works for at least three images and doesn't imply redoing all statements. BTW a person isn't necessarily a human.
- Oppose does work --Pasleim (talk) 21:03, 29 July 2018 (UTC)
- Oppose One usecase for this property (especially when we get structured data for Commons) is when you use a face recognition algorithm to automatically detect the number of faces on an image. You want to source those statements (to indicate that an algorithm is the source instead of a human), but that can easily be done with a reference. And Andy's proposal is a lot more flexible: you can not just indicate how many people there are depicted, but also what gender, age, etc. Here's a SPARQL query showing images that have depicts (P180) with human (Q5) and a quantity (P1114) qualifier. This could better to also take into account subclasses (such as 'woman' or 'man'), but i'm lazy. Husky (talk) 21:11, 30 July 2018 (UTC)
- @Husky: could you give it a try? Just affirming what should be used, isn't particularly helpful if one doesn't even know if it works. Personally, I did try and my attempts on just three of the sample items above easily timed-out. Obviously, we could do anything with a single property per datatype, but computing power isn't infinite which is why we create additional properties. The proposal here is for Wikidata and not Commons. Do you see a disadvantage of having this one if you can just use whatever other solution you prefer for anything else?
--- Jura 15:08, 1 August 2018 (UTC)
- @Husky: could you give it a try? Just affirming what should be used, isn't particularly helpful if one doesn't even know if it works. Personally, I did try and my attempts on just three of the sample items above easily timed-out. Obviously, we could do anything with a single property per datatype, but computing power isn't infinite which is why we create additional properties. The proposal here is for Wikidata and not Commons. Do you see a disadvantage of having this one if you can just use whatever other solution you prefer for anything else?
- Oppose -- JakobVoss (talk) 21:07, 29 October 2018 (UTC)
Not done No consensus.--Micru (talk) 23:05, 8 November 2018 (UTC)