Wikipedia talk:WikiProject Citation cleanup/Repairing algorithmically generated citations

Please add any ideas

I think restricting automated references to a certain permission level is a non-starter, but it could be included. I've included mostly my own ideas and understandings, and probably missed a bunch more. I'm not sure where to discuss e.g. individual failure states on a per-website basis, but future cleanup "worksheets" (like User:XOR'easter/sandbox/ReferenceExpander) can probably be made subpages. Hopefully we'll be able to make better use of tracking categories in the future. Folly Mox (talk) 18:34, 15 June 2023 (UTC)

Community Wishlist proposal

Hi all,

So excited to find this page. Started working on a community wishlist proposal asking for help to improve automatic citations and figured this group may have more to say about that and what can be done. It looks like I need to find a way to improve Zotero based on reading this article? Superb Owl (talk) 22:42, 2 August 2024 (UTC)

Apparently part of the problem that you've been documenting at your community wishlist proposal has to do with how authorship attribution is served on target websites. According to MVolz – the sole maintainer of the Citoid library – some websites only get around to displaying authors once JavaScript is executed by the client browser, so although a particular Zotero library may function for that site in its browser extension form, it will fail to return a parameter when the page is scraped by a crawler.

I might also have misunderstood what MVolz was saying in her comment from last year. Having a look at the project board for Citoid, it seems the Foundation is more interested in integrating Citoid's existing errors into more projects, rather than improving its output.

I haven't really looked into this much since summer 2023, and pivoted most of my citation gnoming to cleaning up after User:Citation bot. It's possible that improving Zotero will lead to improvements in citations generated by the VisualEditor et al, and new translators for domains using their generic parser should certainly help. It's also necessary for basic error checking to be implemented downstream: looking for things like |first2=August 3, |title=Page not found, and the other sorts of nonsense that regularly come out of the algorithms.

The project page attached to this talkpage describes the underlying mechanics of automated citation generation as I (primary contributor to the page) understood them last year. I'm not sure how many other people have looked into the technicalities of this in-depth, and if no one else has better answers for you here (fewer than 30 watchlisters), you might want to ask directly at mw:Talk:Citoid. I certainly hope the Community Tech Team, in their reimagining of the Community Wishlist, have recanted their previous policy of refusing to touch any codebase that is maintained by a different team, although mw:User:MVolz (WMF) is a contractor, so I'm not sure how that works. Folly Mox (talk) 10:09, 3 August 2024 (UTC)