User talk:Rhododendrites (WMF)/Suggested Edits

Feedback, thoughts, and questions welcome

Thanks for taking the time to look through this material (I know there's quite a bit of it). I'd be happy to hear and feedback, thoughts, or questions about the findings or about Suggested Edits in general. If I don't know the answer, I can try my best to find out. --Rhododendrites (WMF) (talk) 18:17, 10 June 2021 (UTC)

A few pings for people who were involved in this village pump thread but may not have seen this yet: @1234qwer1234qwer4, Fæ, C.Suthorn, and Zache: --Rhododendrites (WMF) (talk) 15:20, 15 June 2021 (UTC)

• It's cool to see you did this research, Rhododendrites! I left some of my feedback previously at this thread, some of which has since been addressed, but we have a few shared points. As far as structured data on Commons more generally, my experience so far is that there is a lot of overlap with the category system. Sometimes I add both when I really want to be complete, but other times I think "I just added this photo to the category for the Empire State Building, I really shouldn't need to also say that it depicts the Empire State Building". Ultimately, I hope the systems will be unforked to eliminate the redundancy, which is what I think you've talked about before with categories ultimately being replaced by Wikidata. {{u|Sdkb}} ^talk 18:29, 15 June 2021 (UTC)

• Thanks, Sdkb. I feel like a disclaimer is needed that what I said about categories being replaced by Wikidata was in a volunteer capacity. :) It was also on the English Wikipedia (if I recall correctly), where users and readers engage (or don't engage) with the category system in ways that are different and less essential than on Commons. But yes, there is going to be considerable overlap for the time being. That overlap isn't just with depicts, but also other properties like location, camera model, etc. Some of those are already in place, some are added automatically, and others are still down the road. It might even be worth thinking about early stages of implementing structured data on Commons as, well, organizing the huge amount of information we currently store in a very large hierarchy (not to mention descriptions and filenames) and making that information more readily queryable. --Rhododendrites (WMF) (talk) 12:38, 17 June 2021 (UTC)

What was your methodology?

• How did you determine your sample?
  • In particular, for captions, was this restricted to captions in a particular set of languages?
• What were your criteria to evaluate whether edits were constructive?
  • Is there any chance that you can publish, say, 100 examples of particular edits and your evaluations so that others can decide whether they agree with your estimation of what is and is not constructive? I realize that if you were not tracking as you go, then that is probably asking you to do too much work; in that case, even a dozen or so typical examples would be useful.

Jmabel ! talk 18:56, 11 June 2021 (UTC)

Hi Jmabel. Thanks for taking a look. Did you have a chance to click through to User:Rhododendrites (WMF)/Suggested Edits/data? The sampling, rubric, etc. is documented there (and its two subpages for depicts and captions). But I'm happy to answer other specifics if that doesn't sufficiently cover it. --Rhododendrites (WMF) (talk) 22:46, 11 June 2021 (UTC)

So, looking through this, would it be fair to say that well more than half of "depicts" were more or less accurate as far as they went, but well less than half were even close to being the best choice? - Jmabel ! talk 03:32, 12 June 2021 (UTC)

@Jmabel: Because defining what constitutes "the best [set of] choice[s]" is a little tricky, I'd reframe and say that "while more than half of depicts statements were more or less accurate, most of them were not as good as they should/could be, considering the information Commons has about the files." So, for example, if a picture of a bird has a description of "large white bird," something like "bird" is about as good as we can hope for. If the same file is in Category:Ardea alba, the depicts statement should clearly include "Ardea alba" (or "great egret"), and if the user just tags it as "bird", that's not inaccurate but not as good as it should/could be. But, if it's in Category:Ardea alba but we don't display the categories, it's hard to fault the user for not knowing the bird species by sight. That's the case with the app feature as it currently stands: no categories are shown. This project is to look at the impact of this feature on Commons, but also to consider when the issue is the user and when it's the interface, and what can be done to help one or the other. So among the recommendations for the developers is to make the categories prominent (it might be nice to even prompt the user based on categories, but that's another level of complexity). Another recommendation is to include more specific instructions for what kind of statement is expected (the "tagging" that's common on the internet doesn't always mean "something that you can see in the picture" for example). Between those two changes, I think we can address a good amount of the sub-par statements. --Rhododendrites (WMF) (talk) 15:11, 15 June 2021 (UTC)

I would be very much in favor of making categories readily available in the app. My feeling is that it is crippled without that. - Jmabel ! talk 15:24, 15 June 2021 (UTC)