[01:41:00] Can you share a link for the action constructor? (re @Feeglgeef: Isn't this effectively the action constructor but less broad for no real reason?) [02:09:31] It's a theoretical one; it's almost certain to exist as soon as it's possible. (re @pharosofalexandria: Can you share a link for the action constructor?) [02:10:23] If you'd like to make a specific one on WF with Lexemes we can do that [02:26:14] Here's *a* resignation sentence in Norwegian—which certainly is improvable (I guess it should be 'presidentskapet'?). Note the date provided, and that 'gikk' is the past tense of 'gå'. : https://tools-static.wmflabs.org/bridgebot/3151d23a/file_67154.jpg [02:32:39] Are the languages shown the ones with regular-enough rules? (re @mahir256: Here's *a* resignation sentence in Norwegian—which certainly is improvable (I guess it should be 'presidentskapet'?). Note the d...) [02:33:33] no the languages shown are just the ones I put some amount of time into [02:33:34] only Norwegian, Persian, and Turkish have handling for arbitrary verbs; the others are only good for rendering individual concepts [02:53:02] Ah ok [02:56:38] One more question: since our licensing for the stuff before the abstract content is quite lenient (CC0 for data, Apache 2.0 for code), might some effort be put into making Abstract Wikipedia easy to imbed? It'd be nice if I could: [02:56:39] import abstractwikipedia [02:56:40] abstractwikipedia({some abstract content}).in("English") [03:03:02] that would be pretty nice; in the meantime there's also an endpoint you can send abstract content to via POST requests to get rendered output : https://tools-static.wmflabs.org/bridgebot/4c9c7c00/file_67155.jpg [03:26:14] Yeah. Only issue is that the extra cost effort isn't worth it to most; if I'm going to try to make my service multilingual, the best approach for most people is to use an autotranslate stuff in efficiency/user languages supported well (re @mahir256: that would be pretty nice; in the meantime there's also an endpoint you can send abstract content to via POST [03:26:14] requests to get re...) [03:27:18] My hope is that projects can have full multilingualism with an easy import with just a little translation. [03:29:11] I worry that when we have Abstract Wikipedia fully working we will be very much held back by Wikidata. (re @vrandecic: Newsletter 183 is out! It is again stuffed, and I particularly recommend it! [03:29:12] Contents: [03:29:13] * Sketching a path to Abstract Wikipedi...) [03:30:50] And, I disagree with your stage 4 as a whole. I think that article creators should be explicit about what comes out, just not language specific. [03:32:26] And, as I've stated, I'd prefer the constructor not be "Food with origin and ingredients" but "Description" [03:34:14] Item with origin and components might be a function that can be conditionally called by the Description constructor, but I think a super long list will be hard to memorize and navigate for article creators. [06:51:00] I actually needed the structure that way as it was in the implementation code because it's supposed to give a skeletal structure for an article (re @Feeglgeef: I've cleaned up your implementation. It works now. You were trying to use markdown and include an example usage in it (so, proba...) [07:01:14] Nice example! Note that Waakye was wrongly modeled using P527 instead of P186 (I have fixed it now). (re @vrandecic: Newsletter 183 is out! It is again stuffed, and I particularly recommend it! [07:01:15] Contents: [07:01:16] * Sketching a path to Abstract Wikipedi...) [07:59:22] I can't make my mind up about that, and I am glad it is not up to me but to the wider community 🙂 I also like explicitness, but something like the population number on the other hand shouldn't be repeated, I think. But then, there's a certain advantage to have a single call that produces a lot of text based on Wikidata. It's a give and take. (re @Feeglgeef: And, I [07:59:22] disagree wit [07:59:22] h your stage 4 as a whole. I think that article creators should be explicit about what comes out, just not la...) [08:03:25] That only works for bespoke articles, right? Or how do you imagine it would work if you want to make a batch of articles (like all dishes)? (re @Feeglgeef: And, I disagree with your stage 4 as a whole. I think that article creators should be explicit about what comes out, just not la...) [08:03:53] wikidata tried that and it caused so many issues that we're now in the process of undoing it all (re @Feeglgeef: All languages) [08:05:10] "undoing" or "partially undoing"? I haven't been following in detail. [08:05:38] Through the `mul` label (re @vrandecic: "undoing" or "partially undoing"? I haven't been following in detail.) [08:10:20] partially. we're not removing *all* labels 😅 but copying the same thing to hundreds of languages makes items huge, and leads to huge numbers of edits as people try to mass-add labels for their language [08:11:52] Aah, I misunderstood. I thought you were saying the mul-rollout is being undone, but it is the "name all the things in all languages" which is being undone by the mul-rollout [08:15:01] https://m.wikidata.org/wiki/User:ASarabadani_(WMF)/Growth_of_databases_of_Wikidata talks about some of the issues on the mediawiki database side specifically [08:31:24] I was referring to copying strings that are the same, in particular, but mass-translation of labels causes similar problems because people are still editing each page one by one... I could see people mass-translating labels for implementations or tests, for example, because they tend to follow a pattern (re @vrandecic: Aah, I misunderstood. I thought you were saying [08:31:25] the mul-rollo [08:31:25] ut is being undone, but it is the "name all the things in all langu...) [08:39:21] to be clear, I do care about things being translated, but we should learn from wikidata's mistakes, and a find a better way to handle translations that can be generated by a function (especially given what the wiki is all about 😆) [10:51:23] Agree! [12:57:26] You... don't. (re @Jan_ainali: That only works for bespoke articles, right? Or how do you imagine it would work if you want to make a batch of articles (like a...) [13:06:56] That's one of the biggest wins of Abstract Wikipedia, that we don't need to create each article individually. (re @Feeglgeef: You... don't.) [13:08:12] Sure, but I think that it's important that we construct them manually (re @Jan_ainali: That's one of the biggest wins of Abstract Wikipedia, that we don't need to create each article individually.) [13:08:28] Otherwise, this is not much better than an LLM [13:10:20] I am not sure what you mean or if we really disagree now. (re @Feeglgeef: Sure, but I think that it's important that we construct them manually) [13:16:30] When you say "manually" do you mean only bespoke generated articles (essentially one function for each article to generate), or some other part in a "template function" that would be manual? [14:28:15] I'm away from a computer for this month, but reading the newsletter and listening to the volunteer's corner I have some suggestions for functions someone might like to take up: [14:28:15] * Join list of strings [14:28:16] * Label implementation (supply implementation zid and language, and receive something like "add natural numbers, Python") [14:28:18] * Object to string (with whatever limitations are required), supply language and object, then select the right renderer. [14:28:19] * Label test: supply test zid and language. Receive string of arguments mapping to output (e.g. "[1,2]->3"). Some tests with long input would need the result to be chosen from different levels of verbosity to fit within the character limit. [14:52:51] for the majority of cases, I think projects would probably prefer to describe how an article about a specific type of concept should look, rather than each individual article being converted into a (supposedly) language-neutral form and every language getting exactly the same version of the article, just translated. different things matter to different cultures and the [14:52:51] latter sou [14:52:52] nds like it would just impose an anglocentric view of the topic on everyone else [14:58:48] In the long run I agree, but it could take a few extra decades for every language to write their templates for every different type of object. I think both approaches will be valuable over different timescales. (re @Nikki: for the majority of cases, I think projects would probably prefer to describe how an article about a specific type of concept sh...) [15:15:09] maybe, if that's what the speakers of the language want. we need to be careful about doing things without checking what the speakers of the language want, though, because it could backfire horribly. *we* might think we're doing a good thing, but that doesn't mean the speakers of those languages will see it the same way. converting "our" information into "their" [15:15:09] language could eas [15:15:10] ily come across as another round of colonialism that they want nothing to do with [16:17:12] Hasn't it always been emphasized that is the local projects that will decide what of the available abstract content that they would like to transclude? If no one will get any content unless they explicitly request it we avoid the worst of the problems you mentioned. (re @Nikki: maybe, if that's what the speakers of the language want. we need to be careful about [16:17:13] doing things witho [16:17:13] ut checking what the spea...) [21:45:59] Why do you think so? Just because they have the same content does not mean they are anglocentric. When I "read" the abstract content in the three live languages I'm familiar with, it all makes sense. (re @Nikki: for the majority of cases, I think projects would probably prefer to describe how an article about a specific type of concept sh...) [21:46:50] Yes, this. (re @Jan_ainali: Hasn't it always been emphasized that is the local projects that will decide what of the available abstract content that they wo...) [21:48:04] 1st one exists (re @Toby: I'm away from a computer for this month, but reading the newsletter and listening to the volunteer's corner I have some suggesti...) [21:50:33] I mean, we can have "templates" through compositions of constructors, sure (re @Jan_ainali: When you say "manually" do you mean only bespoke generated articles (essentially one function for each article to generate), or ...) [22:28:16] I've done Earth instead, as the lead for the battle of Alepo has changed significantly and I'd like to make it easy to compare to. It currently has the first full paragraph and I plan to expand it to the whole lead if not the whole article. (re @vrandecic: I'd love to see your abstractification of the Battle of Aleppo article. [22:28:16] About the more fine-grained names of keys, I am not so ...) [22:29:05] [[User:Feeglgeef/Abstract articles/Earth]] [22:34:15] because the people working on this have a strong bias towards english and western ways of thinking about the world. they will include or emphasise things which seem important in western cultures and leave out things which don't seem important in western cultures (re @Feeglgeef: Why do you think so? Just because they have the same content does not mean they are [22:34:16] anglocentric. When [22:34:16] I "read" the abstract con...) [22:36:30] So the people from those cultures can contribute? You don't have to participate in Wikifunctions nor speak English to create abstract content. (re @Nikki: because the people working on this have a strong bias towards english and western ways of thinking about the world. they will in...) [22:40:18] If it's very important, certain Wikipedias can replace the constructors called in the abstract content with ones more suited to their culture, or just manually add it as text on the Wikipedia. [22:42:02] One can imagine the constructor "globglogabgalabian BLP" which templates out a BLP in a way that works for globglogabgalabian [22:42:55] that will still be biased towards western anglocentric views of the world, because that's who has the most power in wikimedia [22:43:39] what is a blp? [22:43:49] Biography of a Living Person (re @Nikki: what is a blp?) [22:44:36] Distinct because articles about living people are different and because of libel policies [22:47:54] I'm not sure what you mean by power here, but it's not like other cultures or non-english speakers are banned from participating (re @Nikki: that will still be biased towards western anglocentric views of the world, because that's who has the most power in wikimedia) [22:48:15] This is a little political, so it might be best to steer clear of this [22:49:50] a good example I know of from wikidata is the property for blood type. like wikifunctions, wikidata is supposedly multilingual and supposed to be usable by other wikimedia projects, but it took three property proposals with lots of opposition before it was finally created, because the western view was that it's private medical information that shouldn't be included [22:56:46] and if someone makes a function that makes an infobox for a person, it's likely that they wouldn't include blood type, because most editors are from places don't care about it. it's also likely that if someone tries to add it, someone else will try to remove it again, because they don't think it should be included by default