Commons talk:Structured data

Structured Commons

About

Get involved

GLAM and culture

SPARQL

Modeling

Discussion

Welcome to the general talk page of Structured Data on Wikimedia Commons. See also specialized talk pages about data modeling and about Lua.

	SpBot archives all sections tagged with {{Section resolved|1=~~~~}} after 7 days.

SDC for Commons Categories

Hi, there are now a couple of thousand files that use Commons category (P373) in the structured data, see here [1]. I'm not understanding the value add for this replication. The standard arguments to use structured data (accessibility, searchability, queryable, etc.) do not seem to apply for SDC giving the commons categories. For one, there are better and more efficient tools to handle categories. Also Commons Categories are mainly about the hierarchy, which is exactly lost in using structured data. Any concerns in actively discouraging the use of use Commons category (P373) in the structured data? @F. Riedelio: : Most of these seem to be added by you. Could you please share your thoughts? Thanks --Schlurcher (talk) 11:54, 27 July 2023 (UTC)

 Answered

Unfortunately, I don't understand the error message or the help for this. Why does the name of the category on Wikimedia Commons, which contains files for the object, have special problems? If the property P373 has been applied incorrectly by me, it can also be deleted. F. Riedelio • 💬 13:06, 28 July 2023 (UTC)

@F. Riedelio Well, properties for SDC are the same properties as those on Wikidata. Not all properties that are useful on Wikidata are useful for SDC. Commons category (P373) is a very important category on Wikidata, where it is used to connect categories at Commons to topics on Wikipedia and other projects. It just doesn't make much sense to use it for files on Commons.

@Schlurcher Categories and depicts (P180) statements so some degree already duplicate each other, I see no good reason to add another layer to that. Do we have a way to block-list certain properties from being used on media files? There is a bunch of properties that should never be used on media files (things like Wikidata item of this property (P1629), for example). Might make sense to prevent them from e.g. showing up in the search results when clicking "Add Statement" on a file page ... El Grafo (talk) 09:02, 31 July 2023 (UTC)

P373 is entirely pointless nowadays (just follow sitelinks), and uses here really should just be removed. Thanks. Mike Peel (talk) 10:03, 31 July 2023 (UTC)

Thanks! I've included a corresponding table in Commons:Structured data/Modeling/Meta. @El Grafo: I'm not sure if or how it is possible to restrict certain of the properties from use. Maybe others can answer. What I can do is enforce this through my bot, i.e., configure it to remove all such statements if it sees them. --Schlurcher (talk) 11:13, 31 July 2023 (UTC)

@Mike Peel: P373 is pointless on Commons, but in Wikidata it is still useful for a Q-item that has a separate item for categories. Let's you get directly from the Q-item to the relevant Commons category without detouring through either that separate Wikidata item or a Wikipedia article or a Commons gallery page. Probably not that important for a piece of software, but very useful for a human who is trying to navigate by hand. - Jmabel ! talk 16:35, 31 July 2023 (UTC)

@Jmabel: It might be a convenience, but it is duplication, and we don't do that for Wikipedia categories. But definitely not useful here! Thanks. Mike Peel (talk) 20:47, 31 July 2023 (UTC)

@Mike Peel: Wikidata has quite a few conveniences/duplications. Think of has part(s) (P527) and part of (P361), where we are always supposed to use both reciprocally, even though in theory we could have only one and calculate the other. - 15:23, 1 August 2023 (UTC)

I think that if we want to expoce the categories via SDC mediainfo we should do it automatically using code, not manually using bots or by hand. --Zache (talk) 15:55, 1 August 2023 (UTC)

Disambiguation of Wikidata items

Tracked in Phabricator
Task T262142

On Tuesday, I opened a discussion on Wikidata, which is here, In brief, the problem is what I encounter by uploading a lot of photographs of Dutch streets on Commons using the upload wizard. After filling in all necessary fields, I am prompted to fill in P180 ("depicts") for the Structured Commons. In my case, this is the name of the street. If the name is relatively common, in the dropdown menu of 7(?) items I get "my" street does not show up. On Wikidata, I proposed to add (by bot) in the field "also known as" smth like "Streetname, Foocity". This would solve the problem here by typing smth "Streetname, F", which would give me a more targeted list in the dropdown menu. The proposal is still open, but at this point it is clear it is not going to get consensus (pls do not try to influence it, this is not the point of my message here). The only other way to solve the problem I see is from the Commons side - to add a function "more" to the dropdown menu, which would bring the next seven items and so on, so that eventually I would get to the item I need. Would the Commons community support this? If not, I will probably stop filling in P180 and recommend to remove the step from the upload workflow, because it becomes a horrible waste of time with only a minimal benefit. Ymblanter (talk) 16:31, 14 September 2023 (UTC)

• I would have no problem with that. - Jmabel ! talk 18:20, 14 September 2023 (UTC)
• I would have also no problem with adding a “more” option to the dropdown, but I fear it’s out of the control of the Commons community. —Tacsipacsi (talk) 22:34, 15 September 2023 (UTC)
• I made that proposal (on someone else's behalf) in 2020. --Matěj Suchánek (talk) 08:02, 16 September 2023 (UTC)

  Thank you, it looks like we are stuck here and there is no way forward. Ymblanter (talk) 09:19, 16 October 2023 (UTC)

It only shows 7 possibles. You end up having to search on Wikidata. Secretlondon (talk) 23:02, 7 October 2023 (UTC)

Qualifier for inscriptions

Hi! I've seen qualifiers like inscription (P1684) (at depicts (P180)). A language is need to set this qualifier. But what about inscriptions with digits only or unknown language? --XRay 💬 07:52, 16 October 2023 (UTC)

I would say digits would work for any language which uses them (though filling in "depicts" with only digits would be weird). Concerning an unknown language, I am not sure: How are you going to fill in "depicts" with an unknown language? Or in any language you do not speak? How do you know it is not nonsense? Ymblanter (talk) 09:19, 16 October 2023 (UTC)

An unknown language may be present if, for example, only individual letters can be read from an inscription. And please have a look to {{Inscription}}. You can use '?' for an unknown language and '~' for digits. I can't find a similar expression for the qualifiers. --XRay 💬 09:45, 16 October 2023 (UTC)

BTW: At the template inscription a '/' can be used for multilingual inscriptions. --XRay 💬 09:46, 16 October 2023 (UTC)

The statements on d:Property:P1684 already provide an example for an unknown language: use the language code und. (I hope the SDC interface allows that, Wikidata clearly does so.)

The case of digits is trickier. Probably the easiest (if applicable) is to use the language of the work/author; this is the language the author likely pronounced the number while writing it down (if they pronounced it). For example, if the inscription is C. Monet 1896, I’d use French because Monet was French. If there’s no clue, maybe I’d use und (Undetermined) here as well. Another possibility is mul (Multiple). —Tacsipacsi (talk) 16:12, 17 October 2023 (UTC)

The example at P1684 is well known, but SDC does not support it. A language is required. --XRay 💬 17:12, 18 October 2023 (UTC)

und is a valid ISO 639-3 language code. If the SDC software doesn’t support valid ISO 639 codes, it’s broken and should be fixed. (SDC software being broken is neither new nor surprising, though…) —Tacsipacsi (talk) 19:36, 19 October 2023 (UTC)

Oh. I'll try to use it. May be another code is good for digits: zxx. --XRay 💬 14:56, 20 October 2023 (UTC)

I don’t think zxx is appropriate for numbers, they are very much linguistic content. To illustrate: depending on the language, the same number may be written as 3 (English), ٣ (Arabic), ३ (Marathi), ༣ (Lhasa Tibetan), Ⅲ (Latin) etc. —Tacsipacsi (talk) 21:17, 21 October 2023 (UTC)

OK. But two other will help in other cases: und (see above) and mul if there are multiple languages. --XRay 💬 08:15, 22 October 2023 (UTC)

Yes, those are useful indeed, although multiple languages can be split up into multiple inscription (P1684) qualifiers in certain situations (e.g. if the same text is repeated in multiple languages). —Tacsipacsi (talk) 13:16, 22 October 2023 (UTC)

Discussion at Commons:Village pump#Blockers to automated import of structured data

 You are invited to join the discussion at Commons:Village pump#Blockers to automated import of structured data. {{u|Sdkb}} ^talk 21:44, 9 November 2023 (UTC)

No metadata about color scheme

Hi. We recently had one query request over at Wikidata where a contributor wanted to look for P18 statements with black and white images, including monochromatic images such as sepia toned images. Unfortunately it doesn't seem like this is included in the structured data as of yet, but it would be immensely helpful in locating P18 images that could use a modern version. Is adding color (P462) claims on Commons images aceptable? And if so, how would you want it modelled? @Bouzinac: Infrastruktur (talk) 16:03, 16 November 2023 (UTC)

I think color (P462) itself is probably best used to describe an image's contents, as a qualifier to depicts (P180) (e.g. to signify that the depicted bus is red). El Grafo (talk) 16:25, 16 November 2023 (UTC)

@Infrastruktur: genre (P136) with either monochrome photography (Q91079944) or black-and-white photography (Q3381576), the latter being more specific. I'd rather have a property based on photographic technique (Q1439691), but it's just a Q-item without a corresponding property. - Jmabel ! talk 18:22, 16 November 2023 (UTC)

Thanks. That seems like a good way to do it. Infrastruktur (talk) 18:53, 16 November 2023 (UTC)