[00:27:24] Is there a phab to track this? (re @Feeglgeef: I just ran into this actually. Something I've proposed before is a system where functioneers have the option to run a call where...) [00:28:18] I believe so. It might be more generic. It's probably buried in "No current plans" (re @u99of9: Is there a phab to track this?) [00:33:10] T287605 is the generic one, it was created by Denny a few years ago. [00:34:51] Thanks. Subscribed (re @Feeglgeef: T287605 was the one, created by Denny a few years ago.) [00:36:18] I understand why caching is so aggressive, to save resources, but it's annoying that entire calls just will be broken forever or at least until the orchestrator decides to mercifully change its mind. [04:09:29] Done. Now with references and a decent default: : https://tools-static.wmflabs.org/bridgebot/b83a852a/file_80885.jpg [08:59:01] That is excellent. Congratulations. (re @u99of9: Done. Now with references and a decent default:) [09:00:22] This can be a good piece to show during Wikimania. [11:17:37] We also need to decide if we want to work with: [11:17:38] * Multilingual punctuations [11:17:40] * Multilingual prepositions [11:17:41] * The level of granularity of the NLG. [11:19:56] For the third point, we have two choices: [11:19:58] 1. Working on subject, predicate, and object as Wikidata item references. [11:19:59] 2. Working on a new function to describe every predicate. [11:21:16] I think that the community is pushing towards 2, as it is certainly simpler. [11:23:26] If the choice is 2, the Page I work in will no longer be relevant. We can just get rid of it after I implement multilingual punctuations and prepositions before Wikimania. [11:24:41] I just used it! https://abstract.wikipedia.org/wiki/Q922 (re @u99of9: I made Z36218 tonight, which has proven fairly flexible and powerful, even if quite crude at the moment. It is supposed to apply...) [11:54:19] I won't be there in person or presenting, but I hope we can make it even more beautiful by then. It would be especially good to have some more languages configured. (re @Csisc1994: This can be a good piece to show during Wikimania.) [12:07:09] I wonder if an abstract article could be made as input for the Ordia tool so that someone easily could see the gap of missing labels/lexemes and then work to close them. [12:08:23] Have you tried David's tool? It can do something in this regard. (re @Jan_ainali: I wonder if an abstract article could be made as input for the Ordia tool so that someone easily could see the gap of missing la...) [12:14:18] Does it? I am looking at https://abstract-data.toolforge.org/article/Q408 and have no clue which lexemes and/or labels are missing in Swedish. (re @u99of9: Have you tried David's tool? It can do something in this regard.) [12:15:49] Am I missing something really obvious? [12:18:44] I was wondering about the same. [12:18:46] No, you're right. I've papered over most of the hard errors (e.g. calling for absent lexemes) by thoroughly using label fallbacks. (re @Jan_ainali: Am I missing something really obvious?) [12:20:06] But I genuinely consider the absence of language configurations to be more problematic than missing label/lexemes. [12:26:28] Sure, but different people may be suited for different tasks and _someone_ will have to create the lexemes or write the labels in each language at some point in time before the configuration becomes useful for that language. (re @u99of9: But I genuinely consider the absence of language configurations to be more problematic than missing label/lexemes.) [12:30:58] Is there a trick to get an article to show? Australia is just circles looping for me. [12:32:12] Wait. It's coming. (re @Jan_ainali: Is there a trick to get an article to show? Australia is just circles looping for me.) [12:33:25] Even after six minutes? (re @u99of9: Wait. It's coming!) [12:34:03] That sounds long but plausible. (re @Jan_ainali: Even after six minutes?) [12:36:53] About 4 s per circle. 2 s to check the cache, then on the second round an actual calculation, also usually in the thousands of milliseconds. [12:38:15] I created an issue for the idea: https://github.com/fnielsen/ordia/issues/305 (re @Jan_ainali: I wonder if an abstract article could be made as input for the Ordia tool so that someone easily could see the gap of missing la...) [12:38:49] I think you may need the tab in focus too. It may stop rendering if you shift to work on another. [12:40:58] That should make around 13 minutes then. (re @u99of9: About 4 s per circle. 2 s to check the cache, then on the second round an actual calculation, also usually in the thousands of m...) [12:41:29] Ah, now it starts to populate. [12:43:54] Now the "quick" way to check if labels are missing is to read the whole article and pick out the words in English 😂 (re @Jan_ainali: Ah, now it starts to populate.) [12:43:55] Seeing it in practice on something more lengthy, I think we shouldn't have any language fallbacks. [12:45:14] You want error messages everywhere? Then en-au would never even be able to use WDs en labels! (re @Jan_ainali: Seeing it in practice on something more lengthy, I think we shouldn't have any language fallbacks.) [12:45:51] A mishmash of English, Swedish, German, Norwegian in the same article just looks like an error anyway. [12:49:05] Your language configuration can choose not to use any fallback functions if you really want. But the default functions should not produce errors IMO. [12:50:15] I'd rather have errors so that it's easier to see what to fix. : https://tools-static.wmflabs.org/bridgebot/06d3ec0f/file_80894.jpg [12:54:24] I think it should be the other way around. If you can tolerate fallback to English, it should be in your language configuration. (re @u99of9: Your language configuration can choose not to use any fallback functions if you really want. But the default functions should no...) [13:01:19] Is it a bug that the entire article is under Q8776414 or an editorial choice? : https://tools-static.wmflabs.org/bridgebot/656df44a/file_80895.jpg [13:05:06] I have not really understood the interface on the left. I already don't like the options I get for adding, and I'm not sure if any of them actually generate whole new sections like lead paragraph. I also don't see why I care what the QID of "lead paragraph" is. : https://tools-static.wmflabs.org/bridgebot/fb79a9c9/file_80896.jpg [13:08:03] There is a tension between what fixers and what readers want IMO. Remember that currently the error messages do not even point to what should be fixed! (re @Jan_ainali: I'd rather have errors so that it's easier to see what to fix. (rendered in Swedish)) [13:16:56] I think we possibly should create technical hindrances to get an article displayed with sentences in a different language than the client wiki (unless specifically being a proper noun or a quote, but those should be marked as intentional in another language). Only "complete" articles should be shown. [13:16:56] Or at least, it should be a community configuration option so that a language version can decide for themselves whether anything goes or not. (re @u99of9: There is a tension between what fixers and readers want IMO. Remember that currently the error messages do not even point to wha...) [13:20:05] To check that I understand, you want to fallback to anything down to mul, but nothing below? In almost all languages, even the names of our predicates (properties) do not have many languages. This is a task that a single hard-working user could do for all properties in their language. But it does not get done, even with the userbase of Wikidata. I've been advocating [13:20:05] for this for [13:20:07] years (https://x.com/tobyhudson/status/1594860721996980224). Breaking an abstract article to make a point will just deliver empty abstract articles. We are trying to make *something* for a reader who has nothing in their language. (re @Jan_ainali: I think it should be the other way around. If you can tolerate fallback to English, it should be in your language [13:20:07] configuration.) [13:23:00] Well, preferably, I would only want labels for proper nouns (and the likes). Anything else should be lexemes. (re @u99of9: To check that I understand, you want to fallback to anything down to mul, but nothing below? In almost all languages, even the n...) [13:23:03] BTW I agree that the full-English-sentence defaults are poor. Usually defaults can be expressed better with symbols wrapping words from the roughly correct language where available. (re @u99of9: To check that I understand, you want to fallback to anything down to mul, but nothing below? In almost all languages, even the n...) [13:24:06] But I am not thinking that this would be "breaking" the abstract article. We would just be clear that the rendering into a particular language is not complete. [13:25:21] In some sense, shoving another language into the rendering is just as much of a "breaking". [13:26:14] "preferably" is a weasel word in this context until we are much further down the track. Do we or do we not display a label if a lexeme is not available, or if the function has not been written with full complexity? I say yes. (re @Jan_ainali: Well, preferably, I would only want labels for proper nouns (and the likes). Anything else should be lexemes.) [13:27:58] I think we just ran the full circle to why I wanted tools to figure out what lexemes to improve/create. (re @u99of9: "preferably" is a weasel word in this context until we are much further down the track. Do we or do we not display a label if a ...) [13:30:43] True, that tool might be good. But I can tell you now that there are about 8000 property labels missing in sv. Many of them are "proper names" that we will likely need. (re @Jan_ainali: I think we just ran the full circle to why I wanted tools to figure out what lexemes to improve/create.) [13:35:49] Those are mostly exernal IDs that only very rarely would be used in a text anyway, right? (re @u99of9: True, that tool might be good. But I can tell you now that there are about 8000 property labels missing in sv. Many of them are ...) [13:37:35] Yes, many are. I expect we will have an external ID box ("Authority control database") down the bottom when we are allowed external links. (re @Jan_ainali: Those are mostly exernal IDs that only very rarely would be used in a text anyway, right?) [13:41:42] Another conundrum is complex concepts like Q1538989. Perhaps I am not just good enough in linguistics, but I would think that in Swedish this is not one lexeme but two in combination. And for such concepts, shouldn't we create the abstract article with such a division in mind. (And is it one or two lexemes in English?) [13:43:42] Yes, I agree. When we have a function with an adjectival (megadiverse) as a parameter, we should replace this call with Z(item=Australia, adjectival=megadiverse, class=country). (re @Jan_ainali: Another conundrum is complex concepts like Q1538989. Perhaps I am not just good enough in linguistics, but I would think that in...) [13:45:33] otherwise we have to teach a function to break apart QIDs which combine concepts. (re @u99of9: Yes, I agree. When we have a function with an adjectival (megadiverse) as a parameter, we should replace this call with Z(item=A...) [13:51:45] @Csisc1994, as shown above _Z33161 _really is not good enough to connect. I think it should remain disconnected. This is one situation where I think broken output is better than slander! (re @u99of9: We should break it and error it for now, so that AW users are not encouraged by seeing any output. I've added a test to show how...) [13:56:02] I see. (re @wikilinksbot: Z33161 – SPO to article-less fragment bandaid) [13:56:13] Thank you. [14:38:13] I've thought of a compromise which I think could work well. If we get past mul we could return the QID as a string. That might help both fixers and readers. (re @Jan_ainali: I'd rather have errors so that it's easier to see what to fix. (rendered in Swedish)) [14:44:16] Do we have requests from any language communities that would like to display abstract articles that are not fully in their language or is it an assumption that some would be okay with it? My limited insight is from Swedish Wikipedia and there I think there would be a strong consensus to only allow abstract articles that fully render in Swedish. [14:46:20] Good question. I guess nobody knows for sure because they're waiting to see what we can make. (re @Jan_ainali: Do we have requests from any language communities that would like to display abstract articles that are not fully in their langu...) [14:46:22] By the way, do we have a term for the act of choosing to get an abstract article to display on a local wiki? Do we "transclude", "import", "display", or something else? [14:51:39] I hope we do not do anything soon! I don't think we are ready. (re @Jan_ainali: By the way, do we have a term for the act of choosing to get an abstract article to display on a local wiki? Do we "transclude",...) [14:53:27] Yeah unless people from, for example, the pilot languages already stated "it's enough if some of the content is in our language, the rest can be in various fallbacks". But I take your answer that no one raised their hand in that way clearly yet? (re @u99of9: Good question. I guess nobody knows for sure because they're waiting to see what we can make.) [14:54:28] We can invent our terminology before actually doing the thing. I even think that should be a mandatory preceding step :) (re @u99of9: I hope we do not do anything soon! I don't think we are ready.) [14:56:45] I can state clearly that an en-au encyclopaedia will happily permit some fallback to en 😆 [14:59:23] Well, that one is even less ready (https://en-au.wikipedia.org/). :) (re @u99of9: I can state clearly that an en-au encyclopaedia will happily permit some fallback to en 😆) [16:23:54] Some positive response by Finn, but also some more technical questions that I could use help answering. (re @Jan_ainali: I created an issue for the idea: https://github.com/fnielsen/ordia/issues/305) [16:29:35] These are controlled by sysops. We still have four slots so if you want anything to be changed you can bring it up on the project chat. From what I've gathered, it looks like there will eventually be a sections feature built into the UI, I was able to control this by editing source. (re @u99of9: I have not really understood the interface on the left. I already [16:29:35] don't like the opti [16:29:35] ons I get for adding, and I'm not sure if a...) [16:39:38] The first five are the defaults, paragraph from sentences was added after consensus on the project chat. I'm a little skeptical about it, if it's to guide newcomers most of the options aren't great, if it's to provide convenience most of the options aren't used. [16:45:43] the list can be changed on-wiki [17:15:34] Forwarded from wmtelegram_bot: Hello, all! I wanted to bring this bug to the community's attention: https://phabricator.wikimedia.org/T423694 . This bug is ultimately due to a content issue for which the orchestrator doesn't return a helpful error. The error-reporting part of this will be handled separately, but the root cause of T423694 is a content issue [17:15:34] with some funct [17:15:35] ion definitions. Details are in the task in case anyone wants to work on this! [17:23:26] Done? [[Special:Permalink/282004]] (re @Sannita: Hello, all! I wanted to bring this bug to the community's attention: https://phabricator.wikimedia.org/T423694 . This bu...)