[12:51:18] I made this: Z29324 It's a bit technical, but I think it will help construct the inner bit of these Wiktionary tables. [13:23:29] I didn't follow the whole thread but how is it hard? (re @Al: Hmm… selecting the “best” form from a lexeme is just a hard problem. In practice, I would have thought it varies according to th...) [13:42:34] Because we don’t know, in general terms, what makes one form better than another, and we have no general way to specify a preference or to prioritise preferences. For example, we can create a function that says, in effect, if it’s an adjective, prefer the positive form or, at least, avoid the comparative and superlative… but how do we know that’s the rule? And how do we c [13:42:34] [13:42:34] hoose which of multiple “preferred” forms is better…? (re @NicolasVIGNERON: I didn't follow the whole thread but how is it hard?) [13:46:55] mhh, what is the context here? [13:46:55] and could a simple rule be "the uninflected (or less inflected) form" (which in most of the case is the main lemma) (re @Al: Because we don’t know, in general terms, what makes one form better than another, and we have no general way to specify a prefer...) [13:48:34] The context here was the labels in the header of a declension table. (re @NicolasVIGNERON: mhh, what is the context here? [13:48:34] and could a simple rule be "the uninflected (or less inflected) form" (which in most of the case ...) [13:49:58] (It started by me trying out Z28602 mentioned in the newsletter.) (re @Jan_ainali: The context here was the labels in the header of a declension table.) [13:51:33] ok... [13:51:34] why not take the lemma? (re @Jan_ainali: The context here was the labels in the header of a declension table.) [13:52:04] Don't ask me, I didn't create the implementation :) (re @NicolasVIGNERON: ok... [13:52:04] why not take the lemma?) [13:53:43] Because we don’t have a general rule that tells us when to prefer the lemma… 😏 (re @NicolasVIGNERON: ok... [13:53:43] why not take the lemma?) [13:54:42] I guess an easy way to find out is to create an implementation that just uses the lemma and we'll quickly find out which languages it won't work well for. [13:54:43] sorry, can you explain like I'm 5 years old, I think I'm missing something obvious... (re @Al: Because we don’t have a general rule that tells us when to prefer the lemma… 😏) [13:57:09] I can't see a single language where the lemma wouldn't be a good header for a declension label [13:57:10] (there is some weird case like languages with multiple scripts, and sometimes inside Lexemes storing lemmata is not easy but that's upstream Wikifunctions) (re @Jan_ainali: I guess an easy way to find out is to create an implementation that just uses the lemma and we'll quickly find out which languag...) [13:59:59] by nature, the unflected form is an obvious choice for a table of flexion (regardless of the language, the lexical category, etc.) [14:00:09] also, again I may miss something, but do we really need a header for such tables? [14:03:15] I am guessing Italian, since that is the native language of the one who built the connected implementation. (re @NicolasVIGNERON: I can't see a single language where the lemma wouldn't be a good header for a declension label [14:03:16] (there is some weird case like la...) [14:12:05] Really? (re @Jan_ainali: I am guessing Italian, since that is the native language of the one who built the connected implementation.) [14:12:14] Who's that? [14:12:38] Dv103 (re @Sannita: Who's that?) [14:12:45] Huh [14:13:44] but my guess is that he is just building a function to find the uninflected form, this is still useful but looking for the lemma would suffice and be simplier [14:14:00] Yeah, definitely [14:16:16] It’s not about a table header so much as a row or column heading. And the form-selection function is a general one, not specific to table headings. The Italian version requires masculine forms for the case, as I understand it (because “caso” is specified) but I don’t see a noun for the number 🤷‍♂️ (re @NicolasVIGNERON: also, again I may miss something, but do we message> [14:16:16] really need a header for such tables?) [14:16:41] also I noticed that https://it.wiktionary.org/wiki/Katze doesn't have an header ;) [14:17:15] aaaaah, now I see (I was indeed missing something) (re @Al: It’s not about a table header so much as a row or column heading. And the form-selection function is a general one, not specific...) [14:18:33] and indeed, that point is less obvious [14:20:32] not entirely sure but just using the label of the grammatical feature would be best in most cases (but I fear there is a lot of weird cases, including for Breton which *as usual* is a mess for declension) [14:23:08] I think that’s what I suggested… it’s a reversal of the ordinary practice for inflected languages, where a specific lexeme is preferred over a simple label. (re @NicolasVIGNERON: not entirely sure but just using the label of the grammatical feature would be best in most cases (but I fear there is a lot of ...) [14:27:13] yeah, I think I agree, nevermind my remarks then [14:27:43] (and I'll have a deeper look for Breton an other day :/ ) [14:39:58] No, I think it was a fair challenge 😎 (re @NicolasVIGNERON: yeah, I think I agree, nevermind my remarks then) [17:52:42] Actually the first offender was English with *Q131105* for example (re @Jan_ainali: I am guessing Italian, since that is the native language of the one who built the connected implementation.) [17:54:33] The lemma of Q1994931 looks great. (re @dvd_ccc27919: Actually the first offender was English with Q131105 for example) [17:54:59] The lemma of L338683 looks great. (re @dvd_ccc27919: Actually the first offender was English with Q131105 for example) [17:58:41] My origina idea for Z27885 was that it should be specialized for language, but until yesterday I had implemented only a draft generic implementation [17:59:50] Now the current implementation calls Z29315, which uses the language configuration Z29317 [18:01:33] Now I've implemented Z29332 (re @Jan_ainali: The lemma of L338683 looks great.) [18:03:50] The original idea was that for some languages the lemma form is not the ideal one to use as a table header (I don't know any particular language, but I wouldn't be surprised if there is one) [18:05:13] Now languages with particular needs can implement their version in Z29317, and even harcode exceptions [19:08:11] I will see for Breton, I can foresee trouble ;) (re @dvd_ccc27919: Now languages with particular needs can implement their version in Z29317, and even harcode exceptions) [19:09:39] maybe not the the header of column/row, but on how to create and/or fill columns/rows (there is *a lot* of weird cases like https://fr.wiktionary.org/wiki/den#br )