[03:28:16] when the upcoming meeting? I really want to be there hehe (re @Sannita: Hi all, the recording of the last Corner is now available on Commons: https://commons.wikimedia.org/wiki/File:Abstract_Wikipedia...) [03:29:48] Usually the first Monday of the month, but @Sannita will confirm. (re @affandymurad: when the upcoming meeting? I really want to be there hehe) [03:30:20] my last chance is the next month for this year (re @u99of9: Usually the first Monday of the month, but @Sannita will confirm.) [04:48:30] @amire80 and I recorded an editing session. I'm not sure anyone wants to watch us for an hour, but here it is: https://commons.wikimedia.org/wiki/File:Wikifunctions_Hebrew_collaboration_session_1.webm [06:02:46] Oooh, there is a broad issue where test evaulations of brand new content use old cached versions of implementations. (re @Al: Yes, but I got a fresh evaluation when I changed the implementation, and it’s still different in the two modes 🤷‍♂️) [06:04:03] I changed Z53028 before I created Z53416, and as-created it failed by using the previous version of the composition. (re @u99of9: Oooh, there is a broad issue where test evaulations of brand new content use old cached versions of implementations.) [07:45:49] I realised afterwards that I was too quick to dismiss your idea of building definites/plurals on the fly. At the very least, these would be good backups when lexeme forms aren't listed. But if extremely regular, they could be faster as a first line construct. In English we do have functions like Z11089, but because there are a lot of irregularities, they can't be [07:45:49] relied on as a p [07:45:49] rimary method. If you want to go down this route, make them as string-to-string functions (should be easy if regular) and we can build with them next time. (re @u99of9: @amire80 and I recorded an editing session. I'm not sure anyone wants to watch us for an hour, but here it is: https://commons.w...) [10:16:16] I’d say your contributions were some way short of excessive. [10:16:16] Perhaps the errors/warnings discussion needs linking back to the proposal for a single output type for NLG functions. I think we’d agree that we need different views of the result object. It’s natural to think of that in terms of the user’s context, but I’m inclined to suppose that it is ultimately up to the user which view they want to see. In any event, if we restrict o [10:16:16] [10:16:18] ur results to objects of a single type, that needs to be a type that supports multiple views. I think it should also support further elaboration within an AW article (as a wrapped function), and it is this that inclines me towards avoiding premature realization as HTML, since that would force the wrapper to parse the HTML, to some extent. (re @u99of9: Thanks. Sorry to [10:16:18] dominate th [10:16:19] e discussion so much!) [13:07:57] There's a deeper root issue, which is that the whole model of storing forms on Wikidata is not great. It leaves a lot of gray area of what should or shouldn't be a form. Every language community can decide it how to do it for its language, which is good for flexibility and freedom, but hard to coördinate and use in practice. (re @u99of9: I realised afterwards that I [13:07:57] was too quic [13:07:58] k to dismiss your idea of building definites/plurals on the fly. At the very least, the...) [13:09:24] And Wikidata also stores forms in a totally concrete way, which is not necessary in most cases. Most speakers of most languages create most forms on the fly as they speak and write. [13:10:16] I tried suggesting to Wikidata developers back in 2018 or so that it shouldn't be done like that, but they decided differently 🤷🏻‍♂️ [13:12:18] As long as the generation in a language follows regular rules, we can use it as a backup or even primary method. (re @amire80: And Wikidata also stores forms in a totally concrete way, which is not necessary in most cases. Most speakers of most languages ...) [13:13:46] For the specific case that you and I talked about, it's a form called "definite" in Wikidata. Someone uploaded a lot of Hebrew lexemes and forms in 2019 using a bot without consulting anyone, and called them "definite forms", but as a Hebrew linguist, I wouldn't call it a "form" at all. It's not a declined form of a noun; it's a noun and an article, just like "the [13:13:46] capital" in Eng [13:13:48] lish. It's just that in Hebrew, it happens to be written as one word. [13:14:15] I understand how it's useful for a spelling dictionary (from which all those forms were copied in bulk), but I'm not sure that it's right for NLG. [13:14:31] my general take on that for my own projects is currently to (logically speaking) do most specific lookups and then slowly make more general ones. if a specific form is missing construct it from the most specific known form. [13:15:19] (such a loop of lookups may not translate into a literal loop of database queries. a good query might reduce the database load signfificantly here) [13:15:32] I have ideas about reorganizing it, but I'm alone, I get no help from anyone, and I don't have the capacity to do it by myself. [13:23:17] Is the rule that you just add "ה" at the start to turn any noun into a "definite"? (re @amire80: For the specific case that you and I talked about, it's a form called "definite" in Wikidata. Someone uploaded a lot of Hebrew l...) [13:25:25] A bit more complicated than that, but not too complicated. I can probably write a function that covers most if not all cases. [13:26:19] Whenever there's a clear convention in WD it's easy enough to match it on the WF side. When there's a mixture (especially within a language) it gets more difficult. (re @amire80: I have ideas about reorganizing it, but I'm alone, I get no help from anyone, and I don't have the capacity to do it by myself.) [14:09:59] we need to confirm that, but it will be probably on December 14, time to be confirmed (re @affandymurad: when the upcoming meeting? I really want to be there hehe) [14:10:17] we're open to experiment with new timings so now is the time to put it on the table [14:10:49] that would be in the early evening (UTC) again? [14:11:00] it should have been on Dec 7, but the whole team will be away/travelling for an in-person meeting, so probably we'll doing it on Dec 14 [14:11:25] @phschafft not sure at the moment, we're open to suggestions [14:11:52] * phschafft nods. [23:41:33] what timezone are you in? (re @affandymurad: when the upcoming meeting? I really want to be there hehe) [23:52:18] https://time-fids.web.app/ GMT +7 : https://tools-static.wmflabs.org/bridgebot/dc67fd96/file_85242.jpg