Community Wishlist Survey 2019/Archive/Create a tag file system in addition to conventional categories

Random proposal ►◄ Archive The survey has concluded. Here are the results!

Create a tag file system in addition to conventional categories

N Outside the scope of Community Tech

Problem: As described most recently at https://wikimediafoundation.org/2018/10/29/george-oates-conversation, Wikimedia Commons has no system for thematically tagging images, such as "myth, stories, night, spirits, stars, or the night sky". The option of "child-friendly" or similar filter would allow children to explore!

Who would benefit: All image reusers and anyone searching for images.

Proposed solution: System of user created non-hierarchical thematic tags for files, as used by almost all other image repositories, like Flickr.

More comments: Suggestion of a "child-friendly" tag/category/filter may be suited in another proposal?

Phabricator tickets: There are many related past tickets, this has been discussed for at least ten years.
phab: T125273 2016 - "Investigate the options available for creating and adding tags for Commons pictures", still sitting in Triage, though this is about automated tags

Proposer: Fæ (talk) 12:13, 30 October 2018 (UTC)[reply]

Discussion

Comment I think before we proceed any further, it should be explained how these tags will work. If I click on a tag "Trees" attached to an image will it take me to a page in which all images have that tag? I can foresee that creating huge unspecific pages in which it's difficult to find anything will be a nightmare. In which case, what point does it have? Rodhullandemu (talk) 14:56, 30 October 2018 (UTC)[reply]

What you are describing sounds like galleries, but there is no intention to create "pages" like galleries. I'm envisioning the way tags work on Flickr, which could piggyback on the existing was that the search engine currently works on Commons, rendering search results on the fly based on user request. For example this search shows any matches for the words "tree, bird, italy" on Commons. If these were thematic tags, the results would be shown as search engine returns but only for files which users had chosen to tag with all of "tree", "bird", "italy", because those files could illustrate all those themes, or were specifically about those themes. On Flickr you get reasonable matches with a search like this, though it seems to use fuzzy matching and most appear to just have birds in trees but no obvious "italy" tag. --Fæ (talk) 15:18, 30 October 2018 (UTC)[reply]

I think Rodhullandemu nailed it, in terms of framing the problem. In my opinion, this “tag” business is pointless for a repository with 50 million items and any results yielded would be illusory (incidentally, that’s why we dissiminate crowded cats). Of course, the same goes for Flickr, George’s boasting in the interview notwithstanding — the only difference is that Commons is not merely a hosting platform but means to be useful for reusers. Tuvalkin (talk) 18:14, 30 October 2018 (UTC)[reply]

Comment Tags in the form of depicts (P180) properties will be part of c:Commons:Structured data, which should begin to be rolled out within a month. I fully agree with need for tags on Commons but lets work within already planned infrastructure. --Jarekt (talk) 15:27, 30 October 2018 (UTC)[reply]
- Cf. Wikidata considered unable to support hierarchical search in Structured Data for Commons --Nemo 15:46, 30 October 2018 (UTC)[reply]
  - software understanding hierarchical tags (so "poodle" will show in search of "mammals", without a need to tag it as "mammal") would be great, but I thought this discussion was about any tags. Also I see a claim that "Wikidata considered unable to support hierarchical search" to be an over statement. Wikidata software is probably able to support it, if the data was set up with that capability in mind, or if it was cleaned up. Wikidata is a wiki so those issues can be fixed, see w:en:Template:Sofixit.
Comment Commons has multiple categories by theme, like Mythology and Dark skies. --NaBUru38 (talk) 16:04, 30 October 2018 (UTC)[reply]
Fæ why is this not a good option? Gryllida 22:52, 30 October 2018 (UTC)[reply]
I've read the George Oates interview now (I'd have many things to say about it... it's always good to have a reasoned outsider's perspective). It seems the actual problem faced in the specific example was the scarcity of clues on what's the actual syntax/technical method to "tag" an image (it's usually not by adding a parameter in the information template). If we wanted to encourage such drive-by tagging, then, a more direct solution would be to make something like HotCat default (with an understandable interface). Usually our conversations about tagging have focused on how to make things easy to find for a large audience. --Nemo 11:07, 31 October 2018 (UTC)[reply]

Fæ and all: This proposal falls under the [[Structured Data on Commons project, which is currently in progress. Community Tech won't be able to work on it, unfortunately. I'm going to archive this proposal, but check out that project page, and you can talk to that team. Thanks for participating in the survey. -- DannyH (WMF) (talk) 19:57, 15 November 2018 (UTC)[reply]