Community Wishlist Survey 2021/Bots and gadgets/Artificial Intelligence Spam hunter

Random proposal ►◄ Bots and gadgets The survey has concluded. Here are the results!

Artificial Intelligence Spam hunter

Problem: Spam is replacing vandalism as our biggest problem
Who would benefit: Everyone. But probably first the English Wikipedia
Proposed solution: Use the articles and edits that have been deleted as spam to train an Artificial Intelligence bot to identify spam, tag it for deletion or revert it where it is certain it is spam and bring it to human attention where it is less confident
More comments:
Phabricator tickets:
Proposer: WereSpielChequers (talk) 22:07, 29 November 2020 (UTC)[reply]

Discussion

I'm a little fuzzy on the distinction between spam and vandalism. Aren't edits that are reverted as spam considered vandalism and used to train ClueBotNG? Is this just about pages rather than edits, then? {{u|Sdkb}} ^talk 04:26, 9 December 2020 (UTC)[reply]
Notice some wikis already have a "draftquality" model available in ORES, which allows using the API to detect new pages which would be deleted due to SPAM. For example, this is the prediction ORES API provides for the page created at w:en:Special:PermaLink/988066833. The API can be used by scripts such as User:He7d3r/Tools/DraftAndArticleQualityCore.js (by adapting the config used at ptwiki), to add icons as in this image. Helder 20:59, 10 December 2020 (UTC)[reply]

Voting

Support Waddie96 (talk) 18:36, 8 December 2020 (UTC)[reply]
Support —MarcoAurelio (talk) 20:04, 8 December 2020 (UTC)[reply]
Support CrystallineLeMonde (talk) 20:21, 8 December 2020 (UTC)[reply]
Support N8wilson (talk) 20:30, 8 December 2020 (UTC)[reply]
Support Ponor (talk) 22:24, 8 December 2020 (UTC)[reply]
Support Hanif Al Husaini (talk) 01:53, 9 December 2020 (UTC)[reply]
Support Keepcalmandchill (talk) 02:17, 9 December 2020 (UTC)[reply]
Support Ottawajin (talk) 05:10, 9 December 2020 (UTC)[reply]
Support Xgeorg (talk) 06:52, 9 December 2020 (UTC)[reply]
Support 1Mmarek (talk) 11:35, 9 December 2020 (UTC)[reply]
Support JASpencer (talk) 16:29, 9 December 2020 (UTC)[reply]
Support Lirazelf (talk) 16:40, 9 December 2020 (UTC)[reply]
Support Rafael ^{(stanglavine) msg} 18:48, 9 December 2020 (UTC)[reply]
Support TheAmerikaner (talk) 20:26, 9 December 2020 (UTC)[reply]
Support YuriNikolai (talk) 22:14, 9 December 2020 (UTC)[reply]
Support NMaia (talk) 01:22, 10 December 2020 (UTC)[reply]
Support Would be interested in helping to develop this, and have been planning a similar project to see if something like this would be feasible. JPxG (talk) 05:45, 10 December 2020 (UTC)[reply]
Support. Meiræ 18:49, 10 December 2020 (UTC)[reply]
Support. Helder 20:59, 10 December 2020 (UTC)[reply]
Support Excellent proposal aimed at combating spam, with human evaluation when not very evident. WikiFer ^msg 23:29, 10 December 2020 (UTC)[reply]
Oppose. Humans must give a reason afther this, to delete. BoldLuis (talk) 10:39, 11 December 2020 (UTC)[reply]
Support ArnabSaha (talk) 15:14, 11 December 2020 (UTC)[reply]
Support SD0001 (talk) 20:35, 11 December 2020 (UTC)[reply]
Support Quebec99 (talk) 21:09, 11 December 2020 (UTC)[reply]
Support Wikibenchris (talk) 08:46, 13 December 2020 (UTC)[reply]
Support.--HakanIST (talk) 08:06, 14 December 2020 (UTC)[reply]
Support —Thanks for the fish! ^{talk•contribs} 22:34, 14 December 2020 (UTC)[reply]
Support Would be a great addition to the capabilities in m:User:LiWa3 (which are currently purely statistical). A bot could read from the feeds of LiWa3, or I could possibly hook it into an external process. Dirk Beetstra ^{T C} (en: U, T) 07:42, 15 December 2020 (UTC)[reply]
Support LucySky (talk) 19:52, 15 December 2020 (UTC)[reply]
Support Ææqwerty (talk) 08:53, 16 December 2020 (UTC)[reply]
Support But only this makes sense in terms of what ORES already can or cannot do, I'd rather it wasn't an unrelated standalone tool/bot Base (talk) 19:47, 17 December 2020 (UTC)[reply]
Support Owleksandra (talk) 19:48, 17 December 2020 (UTC)[reply]
Support If this can be done, it would be useful. This might be even more difficult, but detecting shillsockrings (is that a word? Orangemoody and Wiki-PR and the like) would be even more useful. HLHJ (talk) 02:19, 20 December 2020 (UTC)[reply]
Support AdaHephais (talk) 05:47, 20 December 2020 (UTC)[reply]
Support —2d37 (talk) 10:26, 21 December 2020 (UTC)[reply]
Support David1010 (talk) 11:58, 21 December 2020 (UTC)[reply]