Commons talk:Depicts

From Wikimedia Commons, the free media repository
Jump to navigation Jump to search

This continues to mystify me

[edit]

https://commons.wikimedia.org/w/index.php?title=File:River_Meadows_Park_10_-_a_very_big_tree.jpg&diff=512174546&oldid=453109008

Really? How does this "depict" a river? And (when it comes down to it) do we really want to say this "depicts" Washington state? - Jmabel ! talk 15:07, 12 November 2020 (UTC)[reply]

No response three months later. I still don't understand whether the people who are advocating structured data as a means of doing content description consider edits like this good, bad, or indifferent. I personally consider them bad. I don't think it is useful to take a photo where the normal Commons metadata is quite informative and say it "depicts Seattle." But it's not actively wrong, so I'm not going to invert it. I presume this then is counted -- inappropriately, in my view -- as a "successful" edit with the Android Suggested Edits tool. - Jmabel ! talk 22:49, 12 February 2021 (UTC)[reply]
Jmabel, I fully agree with you that the suggested tool give too broad and incorrect suggestions. When users are accepting these blind, as many(/most?) do, this results in these bad depicts statements. HenkvD (talk) 10:53, 13 February 2021 (UTC)[reply]

Another example: https://commons.wikimedia.org/w/index.php?title=File:Picture111Shopped.jpg&curid=77087150&diff=544336955&oldid=361173272 isn't exactly wrong, but it seems to be a very loose notion of what it means to "depict" something. Is this the intent or not? - Jmabel ! talk 11:41, 21 March 2021 (UTC)[reply]

Or this: [1].

Again: no one seems to be clarifying whether this is what they want people to be doing or not. - Jmabel ! talk 15:38, 21 March 2021 (UTC)[reply]

yes i think the computer-aided tagging is a joke too. there've been many complaints: Commons talk:Structured data/Computer-aided tagging/Archive 2020.--RZuo (talk) 15:56, 21 March 2021 (UTC)[reply]
It's really annoying that I can't tell whether the people running this want these edits (which look like junk to me) or whether I should delete them when they show up on my images. I don't want to be seen as edit-warring, so in the absence of an answer I'm leaving most of it alone, but to me it looks useless or worse. - Jmabel ! talk 15:11, 22 March 2021 (UTC)[reply]
@Jmabel: i forgot to mention, your 4th example was done by Commons:ISA Tool not computeraidedtagging.
i think when the tag is wrong like the 1st example, "river" for your forest photo, that should be removed. washington is also a useless tag.
for tags that are too vague, like the 3rd and 4th examples, "electrical engineering", "product", i would replace them with more specific ones or remove them.--RZuo (talk) 18:02, 22 March 2021 (UTC)[reply]
@RZuo: that's all reasonable, and I could do that, but I'm trying to get some sort of clarification of intent out of the people who advocate for adding depicts. Yes, your judgment is not far from mine, but unless I'm mistaken, you are also basically an "outsider" here. It really bothers me that the people who are supposedly wanting this to happen are not providing any clarity and then, as far as I can tell, are counting anything that is not reverted as a "success". And when I say "reverted" I mean that narrowly: I believe that if there is a further edit, and the tag goes away, it is still counted as a success because it was not reversion as such. - Jmabel ! talk 21:07, 22 March 2021 (UTC)[reply]
Commons talk:Structured data is probably a better place for feedback. however, i dont seem to see any developer or wmf ppl responding there either. no idea who's in charge of sdc or computer aided tagging.--RZuo (talk) 22:32, 22 March 2021 (UTC)[reply]

Another example: https://commons.wikimedia.org/w/index.php?title=File:Women_working_in_Seattle_City_Light_office,_1954_(26662554220).jpg&curid=48766575&diff=545866237&oldid=490766302 - Jmabel ! talk 22:17, 24 March 2021 (UTC)[reply]

This is described as "prominently" depicting a woman, a man, a road, a motor car, and a tree. The last two seem not at all prominent to me; even the others, not terribly useful (might be more so if they had an associated count. They don't). - Jmabel ! talk 23:12, 25 March 2021 (UTC)[reply]

And another: File:Community Sanitation Aides field training, Seattle, 1970 (50290865476).jpg. as shown here. Is this considered a good job of using "depicts" or not? - Jmabel ! talk 23:12, 25 March 2021 (UTC)[reply]

I just think of this as added cruft, much like Automobiles facing right and Categories of Colombia by color. Either it will somehow end up useful or it just sit there and do nothing for the rest of time. mr.choppers (talk)-en- 01:23, 26 March 2021 (UTC)[reply]
Bu the way, over four months later, these "depicts" on File:Community Sanitation Aides field training, Seattle, 1970 (50290865476).jpg are unchanged. Am I right in understanding that when we count whether "depicts" are correct, the fact that no one has removed these has them counted as successes? - Jmabel ! talk 15:46, 7 August 2021 (UTC)[reply]

Further unlikely "depicts" on my pictures

[edit]

User:FlyingKangeroo made this edit. Does a picture of one person really depict an organization of which they were part? Seems wrong to me, but, again, I still don't think I've gotten an acceptable explanation of how this depicts property should and shouldn't be used. - Jmabel ! talk 15:35, 6 September 2021 (UTC)[reply]

Right. A picture of someone involved with the Black Panthers does not depict the Black Panther Party. Ideally we'd have an item for Ron Johnson that would have a statement about being a member of the Black Panthers. For the time being, I just replaced it with "civil rights advocate". There was also a depicts statement for Seattle. The trickiest thing about depicts right now, as far as I'm concerned, is its relationship with other kinds of statements. In this case, the picture really doesn't depict Seattle. It's a picture taken in Seattle, so I added a location statement. The hardest question for me is what to do with the Northwest Film Forum. Does it depict that event? Do all pictures taken at an event depict that event, regardless of the subject of the picture? My own take is that depicts should be limited to what's visible in the picture, and thus the event would be best left to a different statement. The thing is, I don't know if that statement type is enabled on Commons (or if perhaps I just don't remember what it is) or if depicts might be the best we have. This particular edit was added with SuggestedEdits, which I think should have some clearer instructions for users soon. — Rhododendrites talk |  16:09, 6 September 2021 (UTC)[reply]
WRT events, it does exist the property "significant event" (P793), and that property is available in Commons Structured Data. Strakhov (talk) 16:30, 6 September 2021 (UTC)[reply]
@Strakhov: On Wikidata anyway, that property is for significant events associated with a subject. So a major natural disaster, major document being signed, expansion, revolution, coup, etc. associated with a country, for example. I have trouble seeing how that would also cover "the event where a photo was taken." — Rhododendrites talk |  16:46, 6 September 2021 (UTC)[reply]
I do not. It seems natural to me the "significant event" of this picture ("Cannes 2018 Star Wars 2.jpg") is this one (Q43119724, 2018 "Cannes Film Festival". Cheers. Strakhov (talk) 16:52, 6 September 2021 (UTC)[reply]
Fair enough. It looks like the only documentation we have about that property is at Commons:Structured data/Modeling/Significant Event. — Rhododendrites talk |  16:54, 6 September 2021 (UTC)[reply]
With respect to events, see also the discussions at Commons_talk:Structured_data/Modeling#Pictures_taken_during_events and Commons_talk:Structured_data/Modeling/Location#Events? Jean-Fred (talk) 13:59, 7 September 2021 (UTC)[reply]

And another: https://commons.wikimedia.org/w/index.php?title=File:Low_tide_at_North_Wind%27s_Weir_-_from_the_footbridge_04.jpg&curid=106515903&diff=590840754&oldid=568790758 Do we really want to say that something depicts (for example) nature (Q7860)? Or even river (Q4022) and the even more general watercourse (Q355304) when we could say Duwamish River (Q751663) or even more specifically North Wind's Weir (Q27557124)? Doesn't the COM:OVERCAT principle apply equally to "Depicts"?0 ATTN User:Yusufdroid who made these edits. - Jmabel ! talk 01:54, 13 September 2021 (UTC)[reply]

Yes, files can have too many "depicts" entries and, in my opinion, excessive use is unfortunately widespread.
In generally: Probably File:Structured Data on Commons - “Depicts” CC spec.pdf can be helpful to find information about using "depicts". GeorgHH • talk   12:23, 13 September 2021 (UTC)[reply]
I would say "nature" is too general, and perhaps even too abstract for a "depicts" statement, but where to draw the line is not clear. One thing is clear, though: Depicts is not categories, and our rules for categories, based on a hierarchical tree structure, shouldn't simply be transferred to structured data. Specifically, we do want some general and specific depicts statements for depicts. Among other considerations is "what combination of data might a user search for/query for which this image might be useful?" If someone doesn't know they're looking for the Duwamish River or if they're looking for a river with power lines depicted and any river will do, we don't want to assume they're going to query "Duwamish River". Wikidata doesn't have the right hierarchies in place to think of "river" as a "parent" of the Duwamish. Maybe they'll figure out some workaround down the road, but it's just a different sort of organizational scheme. I'd say, based on what I know of structured data on Commons, that the specific river as well as "river" is appropriate for allowing the feature to do what it's supposed to do. That said, what about "water"? Or "blue"? That's where it gets fuzzy, and I don't know if we'll get an absolute answer because of the vast range of possible cases. "Might someone search for X, and might that search be satisfied by this image?" is probably a good guide, subjective as it may be. But ultimately, I don't think anything is harmed by having a bunch of general depicts statements -- it just stops being helpful after a point.
Oh, and just in case it's not clear, since my name appears with "WMF" above: I did a research project for them focused on Suggested Edits, but don't work for them any longer and won't claim to speak for structured data on Commons. Just sharing my own understanding of depicts after many conversations with various stakeholders. — Rhododendrites talk |  14:23, 13 September 2021 (UTC)[reply]
  • @Rhododendrites: It would seem to me that instance of (P31) should consistently support "is-a" inheritance, which should be more effective than inheritance in Commons' tree structure (really only graph structure) of hierarchies, which is just a folksonomy and doesn't guarantee an "is-a" relationship. Is the limit on including more specific Q-values in a query simply one of computation time, or is there actually some theoretical issue involved? - Jmabel ! talk 15:52, 13 September 2021 (UTC)[reply]
  • Whether or not it should be easier to infer hierarchical relationships from linked data, my understanding is it's just not the case that it can be done reliably. I can't really speak to the "why" part, but from what I gather it would take either (a) rebuilding the wikidata ontology from the ground up, or (b) figuring out some clever sort of workaround that nobody has figured out yet. — Rhododendrites talk |  15:59, 13 September 2021 (UTC)[reply]
To recap a bit: early in the history (2018/2019) of SDoC, the WMF advocated for "over-tagging" with depicts statements, saying that using Wikidata ontology for search was not possible/too hard − see eg Commons:Structured data/Get involved/Feedback requests/Good coverage. (At least part of the reason why, I think, the computer-aided tagging tool offers broad tags).
This was, I believe, generally not well received by Commons editors nor Wikidata editors (including me).
What User:Mmullie (WMF) wrote a year ago (Special:Diff/476253278) was that they were still trying to make it happen − which was confirmed in Commons:Structured data/Media search/Future (October 2020). In the ticket phab:T258055#6743284, Matthias concluded 8 months ago that indeed entity traversal was very promising.
I don’t know what the current status is and what happened since then − not sure which Phab ticket is tracking this either. I like to think there is still hope :) We can always ping Keegan to know more.
Jean-Fred (talk) 13:39, 14 September 2021 (UTC)[reply]

And again. I can't say specifically that this is wrong, but even if correct it seems to me to be seriously beside the point.

For what it's worth: the reason I keep coming here with this is (1) I don't see any similarly bizarre tagging choices related to my photos made with categories and (2) during the period when "depicts" was initially being discussed, when I pointed out what seemed to me to be discrepancies in what people were advocating, I was told in no uncertain terms that I was not correctly grasping the intention here. I got out of the way for a year or so, hoping that the result would be something clear. Based on what I am seeing on my photos, it is not. Other than very routine "depicts" such as adding the item for a person where there is already a category for that person, I believe the majority of these I see on my photos go against what I would believe to be good practice. - Jmabel ! talk 15:40, 15 September 2021 (UTC)[reply]

And the hits just keep on coming. Seriously? Depicts United States of America (Q30)? @Top News Anime: please read Commons:Depicts if you are going to be adding "depicts" to files. I, for one, certainly think this should be reverted. - Jmabel ! talk 04:49, 14 October 2021 (UTC)[reply]

https://commons.wikimedia.org/w/index.php?title=File:Localitzaci%C3%B3_del_Carxe-ca.png&diff=602364400&oldid=602364398 : Really? (ATTN: User:FogueraC) A language-distribution map depicts the language??? If so, how exactly is depicts anything other than a duplication of categories? - Jmabel ! talk 01:27, 27 October 2021 (UTC)[reply]

Well, yes, I expect that structured data (not only depicts) will some day replace the manual categories, both in Commons and Wikipedia. Nevertheless, in the meanwhile, it would be great if we could get linguistic maps by language (and by region, date, license, etc.) with the query service, because sparql is a lot more flexible than categories. Therefore, the question should not be if this data is a duplication of categories, but how to model this data. Which properties would you use (or create?) to link a "Map of X language in Y region" with X and Y? FogueraC (talk) 18:27, 28 October 2021 (UTC)[reply]
This should maybe be discussed in a separate section, as it is highly relevant to not only all maps displaying data, but also all kinds of other graphs, plots and diagrams. This does not depict Taxonomy as a discipline. This does not depict a witch trial. Don't we have something along the lines of "illustrates" or "has topic"? --El Grafo (talk) 07:16, 29 October 2021 (UTC)[reply]
main subject (P921)? Strakhov (talk) 14:26, 29 October 2021 (UTC)[reply]
So, a map of X language in Y region should have declarations "main subject (P921) X" and "depicts (P180) Y"? FogueraC (talk) 17:11, 10 November 2021 (UTC)[reply]
@FogueraC: meaning main subject (P921) for the language and depicts (P180) for the region? Sounds a lot better to me than what was done here. - Jmabel ! talk 18:32, 10 November 2021 (UTC)[reply]
Yes. FogueraC (talk) 18:33, 10 November 2021 (UTC)[reply]

And another. Assuming this is, indeed, someone following a suggestion, are these considered appropriate suggestions? house (Q3947) is OK, I guess, but how does my 2009 photo of a an 1890 house in Snohomish, Washington depict 1800s (Q39577), a period at which there no white person had set foot here? - Jmabel ! talk 22:52, 13 May 2023 (UTC)[reply]

Yet again

[edit]

DoesFile:Seattle Daily Times news editor quarters - 1900.jpg really 'depict The Seattle Times (Q221718)? And I cannot even imagine the sense in which it depicts 1990 (Q2064). Pinging @Khimji- Jmabel ! talk 18:11, 14 October 2023 (UTC)[reply]

Depicts spam

[edit]

like https://commons.wikimedia.org/w/index.php?title=File%3AMatt_Hsu%27s_Obscure_Orchestra_-_02.jpg&diff=prev&oldid=743351835&diffmode=source these i removed caused the file to appear in the first row of results when i searched for "bedroom" (in another language) simply because it depicted "bedroom pop".--RZuo (talk) 14:39, 24 March 2023 (UTC)[reply]

is depicts considered a replacement for image maps?

[edit]

and should commons:image map resources be archived for historical interest as a subpage? Arlo James Barnes 09:33, 24 July 2023 (UTC)[reply]

From what I gather, I'd say it's not a replacement per se :
-- FoeNyx (talk) 21:38, 7 October 2023 (UTC)[reply]
Actually, ImageAnnotator could be modified to display depicts statements having relative position within image (P2677). Unfortunately, it has a 14-year-old code base, which is outdated in many ways and likely not really understood by anyone who’s currently active on Commons… But maybe one day someone will implement this feature. —Tacsipacsi (talk) 11:31, 8 October 2023 (UTC)[reply]

Again: what are useful "depicts" values, and do we have any guidelines/casebook?

[edit]

https://commons.wikimedia.org/w/index.php?title=File:I-90_Columbia_River_Scenic_Overlook_pano_02.jpg&diff=prev&oldid=851472036 seems useless to me, but I still don't see any clear standard against which to judge it. - Jmabel ! talk 00:02, 14 February 2024 (UTC)[reply]

I agree. This would be better with genre (P136) landscape photography (Q1353984). Ainali (talk) 07:29, 14 February 2024 (UTC)[reply]
Will edit accordingly. - Jmabel ! talk 21:37, 15 February 2024 (UTC)[reply]
ATTN: User:BobbyWillard. - Jmabel ! talk 21:40, 15 February 2024 (UTC)[reply]

And again: https://commons.wikimedia.org/w/index.php?title=File%3AUniversity_Herald_newspaper_office%2C_1919_%28SEATTLE_1093%29.jpg&diff=538451817&oldid=470411611. Is this actually what is wanted? - Jmabel ! talk 19:51, 26 June 2024 (UTC)[reply]

And again: https://commons.wikimedia.org/w/index.php?title=File:1962_Divco_delivery_truck_01.jpg&diff=prev&oldid=901793773. Can we really say that a picture of a truck "depicts" the corporation that manufactured it? I would think not, but this sort of thing keep getting added to my photos. - Jmabel ! talk 21:17, 21 July 2024 (UTC)[reply]

And another today: https://commons.wikimedia.org/w/index.php?title=File:Seattle_-_view_from_South_River_Street_street_end_03.jpg&diff=prev&oldid=914336937. The specific river is completely identifiable, but this "depicts water"? And according to the edit summary, this was "#suggestededit-add-tag 1.0" which I presume means some bot or such is suggesting these dubious edits to end users. - Jmabel ! talk 00:24, 22 August 2024 (UTC)[reply]

I think the main problem is the wording. The original intention of depicts statements was to replace categories but then someone started calling them tags. And tags on platforms like Flickr and YouTube are something totally different to our categories. GPSLeo (talk) 05:45, 22 August 2024 (UTC)[reply]
Retrieved from "https://commons.wikimedia.org/w/index.php?title=Commons_talk:Depicts&oldid=931483710"