Abstract Wikipedia:Project chat/Archives/2026/05

Abstract Wikipedia should use a different mw-wiki-logo than Wikipedia for Vector 2010 skin

Tracked in Phabricator:
Task T424781

Hello friends. I've been asked to start a community conversation about this. I'd like to propose that Abstract Wikipedia create its own logo, so that folks visiting this wiki don't get it mixed up with Wikipedia. Even though this project is under the Wikipedia domain, I think it's pretty unique and it'd make sense to make sure it doesn't get mixed up with enwiki or other Wikipedias by newbies googling for Wikipedia articles. Thoughts? Thanks. Novem Linguae (talk) 12:02, 1 May 2026 (UTC)

Isn't there already one? Are we talking about a logo in a different context? Feeglgeef (talk) 12:25, 1 May 2026 (UTC)

Looks like this is just for the Vector 2010 skin. Example. I'll edit this section and the ticket to clarify that. Novem Linguae (talk) 12:28, 1 May 2026 (UTC)

Ah, thank you, I'd  Support this then. Feeglgeef (talk) 12:31, 1 May 2026 (UTC)

This has been raised before above in #Logo, the same problem appears on Monobook. — rae^5e <talk> 13:45, 1 May 2026 (UTC)

Would you support using the Vector 2022 logo on Vector 2010 and Monobook? That'd be a simple technical solution. Novem Linguae (talk) 14:02, 1 May 2026 (UTC)

Yes, I see no issue with using that for the moment. — rae^5e <talk> 14:04, 1 May 2026 (UTC)

What would it take for AW to catch up to (English) Reasonator? Porting over functions?

The Wikidata tool Reasonator already automatically creates several sentences in English describing human Wikidata items using auto_long_desc.js under a GPL 2.0 free software license without using any large language model or other AI.

Abstract Wikipedia, on the other hand, requires a several step process just to create a lead sentence. Does Resonator have functions in the English section of auto_long_desc.js that Wikifunctions doesn't have at the moment? Is that what's preventing Abstract Wikipedia from auto-generating substantial text (at least in English) based off of Wikidata like Resonator does?

Below is an example of what Resonator can generate from Wikidata without needing human edits the way Abstract Wikipedia does. (Though I omitted some children for the sake of brevity.)

Johann Sebastian Bach was a Saxe-Eisenach-Holy Roman Empire-German composer, organist, harpsichordist, violinist, violist, conductor, choir director, concertmaster, musicologist, music educator, virtuoso, and school teacher.

He was born on March 21, 1685 in Eisenach to Johann Ambrosius Bach and Maria Elisabeth Lämmerhirt.

He studied at St. Michael's School until April 1702. His field of work included Baroque music and classical music. He was chapelmaster from December 1717 until April 1723 and Thomaskantor from 1723 until 1750. He worked for Collegium Musicum, for Saint Blaise, for Leopold, Prince of Anhalt-Köthen, for Johann Ernst III, Duke of Saxe-Weimar from January 1703 until August 1703, for Thomasschule zu Leipzig, and for Bachkirche Arnstadt from August 1703 until 1707.

He married Maria Barbara Bach on October 17, 1707 (married until on July 7, 1720 ) and Anna Magdalena Bach on December 3, 1721 (married until on July 28, 1750 ). His children include Catharina Dorothea Bach, Wilhelm Friedemann Bach, […]

He died of stroke on July 28, 1750 in Leipzig. He was buried at St. Thomas Church.

Lovelano (talk) 18:41, 1 May 2026 (UTC)

Project scope

Might be a stupid question but what articles are/will be allowed on Abstract Wikipedia? How far is it meant to expand? The language Wikipedias have some differing policies, so might not be as simple as copying those. Personally I find the idea of creating an article for any and every Wikidata item really cool and a good baseline for what can have an article, but wouldn't ~120 million abstract articles become unwieldy? If the only requirement is that the article topic has a Wikidata item, then there are many interesting possibilities; one could write about individual dates, Wikidata test items, even Wikimedia disambiguation pages. Have not found where this is explained, if anywhere. This may be up to common sense, but trouble is, couldn't one create an encyclopedically meaningful article on just about anything? Pretty important policy not to have if it's undecided, though is not a problem at the moment. Just wondering. Some helpful person (talk) 23:36, 1 May 2026 (UTC)

This isn't a stupid question and I don't think this is entirely settled. Have you seen Abstract Wikipedia:About? Koavf (talk) 23:57, 1 May 2026 (UTC)

Ah, I forgot about that page. So it is meant to abstract information from existing Wikipedia articles? Makes sense, though it still raises the question of creating new articles. Some helpful person (talk) 00:41, 2 May 2026 (UTC)

At this juncture, I don't think anyone has in mind drafting anything new here as such: this is just responding to existing Wikidata items and the possibility of new Wikipedia articles drafted from them, using Wikifunctions. Koavf (talk) 00:51, 2 May 2026 (UTC)

I don't think we ever get any article an existing wiki does not have. Even if abstract content creation gets as fast as reasonably possible, it will never beat out typing text, so we will never catch up with, say, enwiki or eswiki. Feeglgeef (talk) 01:10, 2 May 2026 (UTC)

It's also a question of policy. There are topics about which things could be said, but which wouldn't meet the baseline notability or other criteria of the big monolingual wikis. Arlo Barnes (talk) 20:22, 2 May 2026 (UTC)

This is true, if we keep d:'s notability policy (which seems like the most natural one) then there are many subjects that we theoretically can talk about that don't meet the GNG and similar guidelines. Feeglgeef (talk) 22:25, 2 May 2026 (UTC)

How many articles in how many languages are actually fully available without errors

This is a response to the call in f:Wikifunctions:Status updates/2026-05-02.

The most important thing to count is already listed there: How many articles in how many languages are actually fully available without errors? It should be not just a count, but a list: I want to see which articles are readable in which languages; which articles are readable in some languages, but not others (which ones?); which articles are not readable in any language; etc.

An even more relevant, but much harder to measure thing is how many of those articles are actually more useful than not having an article at all. There is no article about Boston in many languages, but Q100, which currently says "Boston is the capital of Massachusetts.Boston is the largest city of Massachusetts." is not significantly more useful than nothing at all even if it's fully rendered in another language in which there is no concrete article about it. Amir E. Aharoni (talk) 12:12, 2 May 2026 (UTC)

Perhaps this should have been on the talk page of that status update instead of the project chat? Feeglgeef (talk) 18:25, 2 May 2026 (UTC)

The status update is about two projects and so I can understand if there are comments here. What do you think about creating a status update overview page with links to the status updates in Abstract Wikipedia to offer the possibility to discuss the status updates and especially things related to Abstract Wikipedia there. Hogü-456 (talk) 18:57, 4 May 2026 (UTC)

I don't think that's necessary. You can just include your abstractwiki (as in, the live wiki, not the project) related comments on the page on f:, or, if for some reason that's not possible, as a reply to the newsletter announcement here. The point is, the project chat is crowded enough with topics as it is, ideally it should be more focussed on general matters of discussion than close-ended enquiries about topics that have their own talk page. I find Amire80's use of a new == heading here to be improper. Feeglgeef (talk) 19:32, 4 May 2026 (UTC)

Wikifunctions & Abstract Wikipedia Newsletter #246 is out: Request for input: what should we count for Abstract Wikipedia

There is a new update for Abstract Wikipedia and Wikifunctions. Please, come and read it!

In this issue, we ask you what would be the relevant metrics for Abstract Wikipedia, we discuss our latest news on Composition Language v2, and we take a look at the latest software developments.

Want to catch up with the previous updates? Check our archive!

Enjoy the reading! -- User:Sannita (WMF) (talk) 12:21, 2 May 2026 (UTC)

Most used fragments

The latest status update lists the most used fragments: "f:Z32123 with 1452, followed by f:Z27868 with 736 and f:Z29749 with 650)".

How was this calculated? Amir E. Aharoni (talk) 13:44, 2 May 2026 (UTC)

Again, this should definitely have been on the talk page for that status update. This isn't the everything page, it should be used for discussion relating to the wiki. Feeglgeef (talk) 18:26, 2 May 2026 (UTC)

It's not a question just about the status update, even if it was prompted by it. It's a question about Abstract Wikipedia that can be relevant beyond the status update.

Stop policing how people use discussion pages without a particularly good reason. If you don't have an answer to the actual question, don't write anything. Amir E. Aharoni (talk) 19:38, 2 May 2026 (UTC)

I have to agree with Feeglgeef that this would be better raised on the status update's talk page itself. — rae^5e <talk> 19:42, 2 May 2026 (UTC)

I'm "policing" which discussion pages are being used for the same reason different discussion spaces exist in the first place. And your distinction between "just about" and "prompted by" is irrelevant. If I want to ask someone how they did something, I would ask them first, not go to a general discussion board. This is still the project chat, not a Q&A zone. @DVrandecic (WMF), can you please answer the question? Feeglgeef (talk) 22:22, 2 May 2026 (UTC)

I have a local script that I have run on a local copy. Nothing more complicated than that. Could also be very wrong -- the script hasn't been peer-reviewed or anything. --DVrandecic (WMF) (talk) 08:54, 4 May 2026 (UTC)

Thanks, @DVrandecic (WMF). Can you share the code, please? I'll be happy to review it. Amir E. Aharoni (talk) 15:30, 4 May 2026 (UTC)

Wiktionary integration

I was trying out a few functions recently, such as the new superlative function, and I noticed that every adjective has to be a Wikidata item. For example, I couldn't type in the inputs "Bugatti Veyron", "fast", "Earth" because "fast" as an adjective is not a Wikidata item. This doesn't really make sense, given that Wiktionary already has the word for "fast" and its superlative form in many different languages. Would it be possible to integrate Wiktionary into the function, so the user can type an adjective or verb from Wiktionary instead of having to deal with Wikidata (which consists almost entirely of nouns)? Somepinkdude (talk) 16:49, 6 May 2026 (UTC)

No, sadly. I would try "speed" as the adjective, though, you have to use abstract qualities instead of concrete English terms.

Here's an example of the word being used.

Feeglgeef (talk) 18:02, 6 May 2026 (UTC)

Wikidata has standard items as well as lexeme data, so if anything, we should try to integrate those. E.g. d:Lexeme:L1029. ―Justin (koavf)❤T☮C☺M☯ 19:48, 6 May 2026 (UTC)

I tried this, and it gave "Bugatti Veyron is the speedest car on Earth", which is laughably bad. The problem is that Wikidata does not "know" what the superlative form is, while Wiktionary does. Speed is a fairly common descriptor, but what about less common adjectives which aren't on Wikidata? Will there need to be a bot to copy all of this information from Wiktionary? Somepinkdude (talk) 22:14, 6 May 2026 (UTC)

I'm not suggesting that the functionality works now, but that we need to integrate the funcationality. ―Justin (koavf)❤T☮C☺M☯ 22:17, 6 May 2026 (UTC)

Well, no, the superlative form is chosen by f:Z12203. I think there's a bug with the English implementation that calls that function wrong. Feeglgeef (talk) 22:19, 6 May 2026 (UTC)

@Somepinkdude I did find the item fast (Q19807466) which was linked to a concept database's entry for "fast"; I have at least linked it to the English lexeme "fast" and will try to add more links at a better hour. Mahir256 (talk) 04:18, 7 May 2026 (UTC)

Showcase

I was wondering if there's a curated list of decent articles to learn by example here on Abstract Wikipedia? Fgnievinski (talk) 02:02, 8 May 2026 (UTC)

Unfortunately not, and, honestly, I wouldn't recommend article creation at the moment, it's likely to need significant refactoring as the way we represent abstract content changes, but, I can point you to Q668 as a reasonably good example at the moment. Feeglgeef (talk) 02:39, 8 May 2026 (UTC)

Thank you. "308 languages" blew my mind – well done! Fgnievinski (talk) 02:54, 8 May 2026 (UTC)

308 is the number of concrete Wikipedias in which a manually-written article about India is available.

The abstract article renders for me in English as "India is a country in Asia. India is a republic. New Delhi is the capital of India. India is the most populous country in world." Switching to another language to get rendered as an abstract article is done by typing the language's name in the box above the rendered text. I tried French, German, and Dutch, and got errors in all of them. Amir E. Aharoni (talk) 03:12, 8 May 2026 (UTC)

So it goes. The paragraph has quite a lot of distinct sentence generation functions, all of which of course have to be implemented in the target language. This will be rounded out once we're able to more effectively generalize linguistic content. I view these functions as sort of patchy temporary solutions to the sentence generation problem while we figure out something more robust, which I hope is soon to come. — rae^5e <talk> 04:15, 8 May 2026 (UTC)

Wikifunctions & Abstract Wikipedia Newsletter #247 is out: References from Wikidata now available

There is a new update for Abstract Wikipedia and Wikifunctions. Please, come and read it!

In this issue, we announce that is now possible to pass references in Wikidata statements, we introduce the Abstract Data dashboard, we report you on the presentation about Abstract Wikipedia at WikiCon Australia, and we take a look at the latest software developments.

Want to catch up with the previous updates? Check our archive!

Also, we remind you that if you have questions or ideas to discuss, the next Volunteers' Corner will be held on May 11, at 17:30 UTC (link to the meeting).

Enjoy the reading! -- User:Sannita (WMF) (talk) 11:16, 8 May 2026 (UTC)

about the usage of Z33068

The use of f:Z33068 (en: paragraph from sentences) always most of the time results in the error "Reached time limit in orchestrator." So should it be used at all for now, temporarily? Or maybe there are already some tickets about this that I am not aware of? -- Asked42 (talk) 18:23, 10 May 2026 (UTC)

It previously worked, so I'm hoping whatever the issue is will be fixed soon. The current state of the wiki makes immediatism an illogical philosophy to have at the moment, in my opinion. Feeglgeef (talk) 20:16, 10 May 2026 (UTC)

This issue seems to no longer exist, you might need to clear the cache by making a dummy edit on affected articles. I implemented the function in Python (instead of composition), which has increased the speed by about 2000 ms. Feeglgeef (talk) 04:31, 11 May 2026 (UTC)

Seems like working. Thank you. -- Asked42 (talk) 19:06, 11 May 2026 (UTC)

Manual of Style

I've created a draft for a Manual of Style. Please feel free to comment on or make bold edits to the page. Feeglgeef (talk) 03:29, 11 May 2026 (UTC)

Thank you! That’s great and I’ll look at it more soon. I’m still taking a bit of a wikibreak. Immanuelle (talk) 22:32, 14 May 2026 (UTC)

Is there anything I've written here that you think would make for a good technical section of the page? — rae^5e <talk> 00:04, 15 May 2026 (UTC)

I've added some of them. Thank you for your list of best practices by the way, I generally agree with them (I can't endorse connecting WD items currently in hope of a better technical solution, but otherwise I do). Feeglgeef (talk) 00:58, 15 May 2026 (UTC)

Proposals on the architecture of Abstract Content rendering

Starting from a discussion born on the Wikifunctions Telegram chat, I've explained two different proposals on how the NLG on Abstract Wikipedia should be organized in the page User:Dv103/Abstract articles architectures. Please come to contribute to the discussion, or to propose alternatives. Dv103 (talk) 14:32, 11 May 2026 (UTC)

MediaWiki:Wikimedia-copyright-footer

The copyright message has a few issues, mainly that it does not address abstract content (which is not "text"), nor object labels (which are under CC0). Proposed text:

" Text and abstract content are available under the [https://creativecommons.org/licenses/by-sa/4.0/ Creative Commons Attribution-ShareAlike License]; the labels of objects from [[f:|Wikifunctions]] are available under the [https://creativecommons.org/publicdomain/zero/1.0/ Creative Commons CC0 License]; additional terms may apply. See [https://foundation.wikimedia.org/wiki/Special:MyLanguage/Policy:Terms_of_Use Terms of Use] for details. "

Which, when rendered, looks like:

Text and abstract content are available under the Creative Commons Attribution-ShareAlike License; the labels of objects from Wikifunctions are available under the Creative Commons CC0 License; additional terms may apply. See Terms of Use for details.

The message is somewhat inspired by the current one and WF's. Thoughts? Thank you! Feeglgeef (talk) 13:30, 12 May 2026 (UTC)

@Feeglgeef: As you probably imagine, this text is very tightly controlled by the Wikimedia Foundation's Legal department. I think the first port of call should be a discussion about why you might want to change things, rather than a rush to specific re-wordings, before approaching them for sign-off. Jdforrester (WMF) (talk) 14:28, 12 May 2026 (UTC)

Well I briefly summarized my justification, my proposed text was just an idea for a starting point. Basically, the abstract content is not really "text", so the footer doesn't actually tell the reader what they're allowed to do with it, and I argue that the current wording implies that the labels are licensed under CC-BY-SA 4, when they aren't. I think a change is critical as otherwise we're not showing the reader a license, or, arguably, in the case of labels, showing users a license that does not apply, which is a pretty bad thing. Feeglgeef (talk) 14:51, 12 May 2026 (UTC)

Abstract Wikipedia statistics site source code

https://abstract-data.toolforge.org/ is a very useful tool, but it doesn't seem to have anything pointing to the location of its source code. Is it closed-source or unfree?

While I'm at it I feel I should suggest having language pages' paths formatted differently from how they are currently—at the moment it's /languages/Toki%20Pona when something like /languages/tok would be more suitable. — rae^5e <talk>

See gitlab:toolforge-repos/abstract-data. Also pinging @DSantamaria-WMF (thanks for the tool, by the way). NguoiDungKhongDinhDanh 00:25, 13 May 2026 (UTC)

Thank you. I think a link to that repository should be put in a footer somewhere on the website as I wouldn't have found that if not for you. — rae^5e <talk> 02:30, 13 May 2026 (UTC)

@Theki: For the record, I found that repo by searching for "abstract-data" on Toolhub. NguoiDungKhongDinhDanh 03:34, 13 May 2026 (UTC)

Thanks for the feedback and the report, fixed both things, we have now the link to the repo in the side bar footer, and also the language codes are used in the URLs. Thanks! DSantamaria-WMF (talk) 11:46, 13 May 2026 (UTC)

Are you particularly attached to keeping country flag emojis next to the language names? Some consider it an anti-pattern. Arlo Barnes (talk) 16:17, 14 May 2026 (UTC)

I think the use of flags is not really harmful since this is not a language selector, it is just a list of languages. Of course not every language has an associated flag, for example Latin, which does look a bit awkward, but it can't be helped. Maybe give languages without a flag a unique icon from Wikimedia Commons? e.g.

Toki Pona

instead of

🌐Toki Pona

. — rae^5e <talk> 16:37, 14 May 2026 (UTC)

I thought I saw 🏛️ being used for Latin earlier. But of course, as far as the Unicode Consortium is concerned, 🇬🇧 doesn't mean 'English', no more than 🏳️‍🌈 means 'Polari'; technically, it doesn't even mean 'United Kingdom', just the letters 🇺 and 🇰 next to each other (which linguistically would indicate Ukrainian in ISO 639, but these are 'regional indicator symbols'). Arlo Barnes (talk) 17:21, 14 May 2026 (UTC)

As seems (from this comment and also others' comments) that flags are a little bit controversial, I just removed them, IMHO it was just an aesthetic concept, I like the look and feel that flags bring to the language labels, but happy to remove them if make some people feel uncomfortable with the concept. DSantamaria-WMF (talk) 05:54, 15 May 2026 (UTC)

Make it more user Friendly

Make it more user-friendly by adding drag-and-drop compenots and one click additions, and make it similar to a visual macro buidler. ChippyTechGH (talk) 01:00, 15 May 2026 (UTC)

Could you file something like this on phabricator? Feeglgeef (talk) 01:01, 15 May 2026 (UTC)

Person infoboxes

Hi all, I've just created infobox for person (Z35167). All it can really do presently is display the name of the person in the chosen language and in the person's native language, but it's a good start. I started writing it as a composition but chose to switch to writing it in JavaScript instead, which ended up being less headache-inducing (managing a large list of strings to concatenate together is not fun, plus compositions bring lots of additional overhead). Unfortunately, this means that the actual useful parts of the infobox—the information about the person themself—is not really doable. I'd need to dereference the Wikidata property references to get their label in the requested language, but since they're references and I'm working with code, not compositions, I can't just call fetch Wikidata property (Z6822) and be done with it. If there's a WikiLambda API to do this inside of the JS itself that would be great and would solve everything, but at the moment that alone is what's holding it back, or so it seems. I can get the actual values of statements just fine, though. If I am missing some amazing WikiLambda API function that will solve this problem for me by letting me dereference the PID, then please let me know because that would be great.

If you want to look at an example of the infobox right meow, you can view Q317521 in Toki Pona and see how it states Musk's name in both Toki Pona and his native English name. COOL!

My hope is that this can be made less barebones in the near future. Infoboxes are a great way to get quick information on a subject and can fill in the informational gaps in the presently sparse articles we have at the moment, all while staying in the user's requested language. Feedback is appreciated! — rae^5e <talk> 02:58, 15 May 2026 (UTC)

Thank you for this! I wonder if perhaps it would be better to create one "infobox" function and then have compositions call that, making wd retrival possible? Feeglgeef (talk) 03:38, 15 May 2026 (UTC)

Working on that at the moment, I think I've got it working but I'm hitting a ratelimit in the orchestrator so I don't know for sure. The main issue with it now is that the actual infobox template (Z35175) wrapper was programmed in JS so as to not be a pain in the ass, so this means that it can only take in strings, HTML fragments, etc. so the retrieval of Wikidata stuff still needs to be done by the caller and is rather verbose (see the main implementation). The niceties will come soon later once I know that it all works at a lower level, though. — rae^5e <talk> 05:04, 15 May 2026 (UTC)

So the problem now seems to be that when label of Wikidata property in language (Z29825) isn't being faulty, it's being incredibly slow. Considering retrieving Wikidata items from their references seems to take, like, four to five seconds, and we'll have tens of statements in these infoboxes, I don't see how the property names could be retrieved in a way that doesn't cause the orchestrator to time out or hit a ratelimit. — rae^5e <talk> 06:01, 15 May 2026 (UTC)

I think I figured something out? I'm just going to try calling fetch Wikidata property or parts (Z35036), it seems to be much faster for whatever reason. — rae^5e <talk> 06:05, 15 May 2026 (UTC)

It works now, a little rough around the edges, also very slow, but I think I'm mostly happy with it. Of course it would be great if all you needed to provide to the template function to get a label-statement row was the property reference itself, but I'm too sleep-deprived to figure that out at the moment. I made five bespoke functions specifically for this infobox, more than I would like, but it's probably not that bad. I'm just hoping I didn't duplicate any functionality in trying to make the composition less verbose and repetitive. — rae^5e <talk> 07:15, 15 May 2026 (UTC)

Wikifunctions & Abstract Wikipedia Newsletter #248 is out: A higher meaning

There is a new update for Abstract Wikipedia and Wikifunctions. Please, come and read it!

In this issue, we discuss functions creating language fragments, we present our latest news in Types, and we take a look at the latest software developments.

Want to catch up with the previous updates? Check our archive!

Enjoy the reading! -- User:Sannita (WMF) (talk) 14:36, 15 May 2026 (UTC)