Wikipedia talk:Wikipedia Signpost/2023-08-31/Recent research

Discuss this story

Really good to see some research in this area. It deserves so much more honestly. —TheDJ (talk • contribs) 09:15, 2 September 2023 (UTC)[reply]

Minor matter, I suspect ". . . host more images of a subject that could conceivably be needed to illustrate pages on other projects . . ." might better be "than could".

Down to substance, somewhat, I sometimes hope Wikidata will help with our problems with searches, categorization, and languages in Commons. Some day, far in the future, I fear. Thus far, the cost (mainly in annoyance to old-time cat wranglers like me) of implementing Structured Data is more easily visible than the benefit. Jim.henderson (talk) 23:53, 2 September 2023 (UTC)[reply]

Fixed, thanks! (the typo, not Commons - yet)

Btw, someone also started a thread on the Commons village pump about this review and paper. Regards, HaeB (talk) 03:36, 4 September 2023 (UTC)[reply]

Excellent research topic. I think a major problem for the future is the new skin not putting a link to Wikimedia Commons on the left. It massively hinders discovery of the resources Wikimedia Commons has, which people should be able to find. I intend to petion more about this in future, and it's a major reason why I've switched back to the old skin.
The point about photo replacement is very good to note. This could be added to the watchlist ("on Commons, a photo on the article was overwritten with a different one").
With the problem that Wikimedia Commons has lots of photos: this is a good thing. People may be looking for something unexpected. There's a trend for major topics on Wikimedia Commons to create a curated highlights gallery of images and link wikipedia articles to this rather than the category. I'm very opposed to this, it hinders discovery of photos if people are looking for something unusual that the gallery creator wasn't expecting. It's much better to have that kind of gallery in the Wikipedia article, presenting photo highlights, and a link to the Wikimedia Commons category for people looking for something unusual. Blythwood (talk) 23:16, 4 September 2023 (UTC)[reply]

The characteristics, or problems, identified all exist, but are perhaps not major. Commons is absolutely enormous, and the "curation" process can only work in a very patchy way. Commercial picture libraries work on the principle that you can never predict what will interest a user, so you just offer everything you have. So I think the solution proposed for "barrier 2" "The process would enable editors from both platforms to figure out whether an image warrants significance in any contexts collaboratively, rather than relying on judgement of editors from one platform or the other" is a complete non-starter. Johnbod (talk) 23:34, 5 September 2023 (UTC)[reply]

- It does feel as if non-English searches get pretty short shrift on Commons. For instance "forest" gets 1,321,948 hits, but "forêt" (in French) gets 29,638. "skogen" ('the forest') in Swedish gets just 1,329, and the first batch of images are about a place with that in its name, not forest at all; whereas "skog" (forest in general, in Swedish) gets 1,459,858, and the first batch combines quite a few images of forest with people of that name. The Greek "δάσος" (forest) gets 1,335,550, but many of the first batch of images are of mangrove forests, which might seem rather a specialised result. My curiosity piqued, I tried the Indonesian for forest, "hutan" (rain forest is "hutan hujan", a good word, huh?): it returns an impressive 1,323,199 results, the first batch mixing forest resorts, forest parks, forest, and a boat named after a forest. On the other hand, "hutan hujan" on the other hand returns 13,228 results, where "rain forest" gives just 5,872. About all one can safely conclude from this brief and wholly unscientific experiment is that search is language-dependent, or to put it another way, not very reliable for anyone arriving and searching in their own language. Chiswick Chap (talk) 17:10, 9 September 2023 (UTC)[reply]