[09:51:01] Hi y'all, [09:51:02] where are we about inputs for NLG functions? [09:51:03] if I followed correctly there is no proper way to do it, and the few prototype are going in a lot of different ways (like number for Z21617 or sign for the gender on Z21765 - which is a bit sexist btw) [10:00:23] It’s certainly a hard problem. I think we need a few anchors, like grammatical features on Wikidata and generalised phrase structures that can apply across multiple languages and vary within a language. Basically, natural-language silos are a dead end. (re @Nicolas: Hi y'all, [10:00:23] where are we about inputs for NLG functions? [10:00:24] if I followed correctly there is no proper way to do it, and the few prot...) [10:07:37] it is indeed hard (re @Al: It’s certainly a hard problem. I think we need a few anchors, like grammatical features on Wikidata and generalised phrase struc...) [10:07:39] not having a proper way to do things means people are creating functions with strange and inconsistent inputs which is even worse... [10:08:05] so we really should think about it and find some kind of solution... [10:10:53] We could begin a dialogue on [[Wikifunctions:Project_chat]] (re @Nicolas: so we really should think about it and find some kind of solution...) [13:56:38] https://www.wikifunctions.org/wiki/Wikifunctions:Project_chat#Natural_language_functions (re @Nicolas: so we really should think about it and find some kind of solution...) [18:12:49] Close enough. It’s just a placeholder for the pronoun, which would be “sie” rather than “es” for “ein Mädchen”, but it might need to be gender-neutral. (Number and case are implicit.) I don’t know how we’d account for the use of “ce” in French, when that’s appropriate. (re @Nicolas: like "Mädchen" which has neuter as grammatical gender but feminine as semantic gender) [19:43:11] yes, it's "sie" for "Mädchen" because pronouns (usually) follow natural gender and not grammatical gender, but the articles do follow the grammatical gender "das Mädchen" and not "*die Mädchen" (re @Al: Close enough. It’s just a placeholder for the pronoun, which would be “sie” rather than “es” for “ein Mädchen”, but it might nee...) [19:46:55] which means that *if* we create several datatypes one for each language, we would need also need several datatype _inside_ some languages [19:46:56] which means creating probably more than 10 000 datatypes just for genders (and at least as much for numbers, cases, cases, etc.) [19:46:58] I doubt the developping team or the community would agree to create or deal with 100 000+ datatypes [19:46:59] one big datatype for all grammatical features seems more logical [19:59:39] I wouldn’t look at it that way. When I say it’s a “placeholder”, that implies a function rather than a type. In this case, the function might return a “noun phrase” type of object whose features could include case, number and gender. The appropriate pronoun is a default in the case where there is no specific noun. (But I’ll expand on this on-wiki.) (re @Nicolas: whi [19:59:39] [19:59:40] ch means that if we create several datatypes one for each language, we would need also need several datatype inside some lang...) [21:42:01] I think we should have a system where functions can have a custom dropdown (re @Nicolas: which means that if we create several datatypes one for each language, we would need also need several datatype inside some lang...) [21:42:49] Those values would convert to whatever, and any other function using the same items would just use the same object [21:43:32] Or, the dropdowns could be a list of references to a "Dropdown item type" for consistency [21:44:23] So basically enums but less restrictive [21:59:36] Is that part of T379338? Would you see it as a list of pairs, Z12 + Z1, where the user interface selects the relevant Z11 from each Z12 to make a list for the user to select from, and the corresponding Z1 becomes the input argument? (re @Feeglgeef: I think we should have a system where functions can have a custom dropdown) [22:00:09] Yeah (re @Al: Is that part of T379338? Would you see it as a list of pairs, Z12 + Z1, where the user interface selects the relevant Z11 from e...) [22:00:21] Or references to pairs for reusability [22:03:26] Wouldn’t that just be a Z2, with the Z12 being its Z2K3? (Or a list of such Z2s for reuse?) (re @Feeglgeef: Or references to pairs for reusability) [22:04:13] Yeah (re @Al: Wouldn’t that just be a Z2, with the Z12 being its Z2K3? (Or a list of such Z2s for reuse?)) [22:56:44] but how do assure consistency between functions? (re @Feeglgeef: I think we should have a system where functions can have a custom dropdown)