[00:09:55] *** Quits: wmacmil (~wmacmil@c83-252-138-144.bredband.comhem.se) (Ping timeout: 240 seconds) [02:39:29] Lycurgus: yes, as a phonetic phenomenon, sandhi etc. is everywhere [02:39:35] would be hard to speak without [02:40:33] just in a purely phonetic level, if you say some consonant (say [k]) between different vowels, it's slightly different every time [02:41:06] when you say [ku], the k you're forming is already affected by the u that's coming next, your lips are prepared to get rounded [02:42:45] in most languages, those are just phonetic differences, not phonemic (https://english.stackexchange.com/questions/104224/what-is-the-difference-between-phonetic-and-phonemic) [02:46:41] re: contractions or just "words", I don't really care about that distinction, I just define things for myself in terms like, "how many GF subtrees do we need to decide on a string of characters separated by spaces" [02:47:08] and that's what linguists usually call morphosyntax :-D [02:47:32] in the most boring level we'd just have each constructor contribute with a string and they're all concatenated [02:48:26] and in the most complicated level, we need to know everything about the sentence (speaker, object, tense, aspect, mood) to produce just one string [02:48:59] so when I run into a thing where adding an adverbial like "in the house" to a sentence like "I saw you" [02:49:12] and that preposition "in" somehow merges with "you" [02:49:27] that's pretty exciting from GF resource grammar writing perspective! [02:49:30] *** Joins: drbean_ (~drbean@TC210-63-209-35.static.apol.com.tw) [02:50:01] I know I'm completely ignoring your monologue about semantics and responding with my own :-D we can call that a conversation [02:53:35] but I am probably just a wimp, because I like these problems which are totally solvable [02:54:01] semantics and pragmatics, I wouldn't dare to say that [03:38:57] *** Quits: drbean_ (~drbean@TC210-63-209-35.static.apol.com.tw) (Ping timeout: 264 seconds) [04:04:10] *** Quits: fl0_id (~fl0_id@claudia.s7t.de) (Ping timeout: 256 seconds) [04:16:38] *** Joins: fl0_id (~fl0_id@2a01:4f8:171:4de::40:2) [05:21:18] < inariksit> and that preposition "in" somehow merges with "you" <-- with this I don't just mean that the strings go together like "in+you" or "you+in", that's still in the boring category, but I mean that they morph into something completely different, so we need to include those forms in the inflection table of either "you" or "in" [06:27:24] *** Joins: wmacmil (~wmacmil@c83-252-138-144.bredband.comhem.se) [06:37:54] *** Quits: wmacmil (~wmacmil@c83-252-138-144.bredband.comhem.se) (Ping timeout: 246 seconds) [07:49:07] *** Joins: wmacmil (~wmacmil@c83-252-138-144.bredband.comhem.se) [09:06:45] *** Joins: Lycurgus (~niemand@cpe-45-46-139-165.buffalo.res.rr.com) [09:10:07] inariksit, funny that's just the kind of think chomsky was on about in the mentioned talk: https://www.youtube.com/watch?v=X4F9NSVVVuw about 1:10 in [09:10:14] *thing [09:32:37] *** Quits: wmacmil (~wmacmil@c83-252-138-144.bredband.comhem.se) (Ping timeout: 276 seconds) [09:47:08] *** Quits: Lycurgus (~niemand@cpe-45-46-139-165.buffalo.res.rr.com) (Quit: Exeunt) [10:22:21] *** Joins: wmacmil (~wmacmil@c83-248-104-92.bredband.comhem.se) [10:56:01] im wondering why i'm getting the following [10:56:01] QueryEngRgl.gf: [10:56:02] Warning: function p is not in abstract [10:56:02] Warning: no linearization of ListFun [10:56:13] when i have a cleary defined linearization of ListFun [10:56:27] and i don't see a function p being called anywhere in my linearization [10:57:49] for reference, this is in https://github.com/wmacmil/AgdaGrammar/tree/main/pgfTests [10:57:54] QueryEngRgl.gf [11:00:17] nvm, there was a typo p hidden i wasn't seeing that was obstructing everything [11:22:02] what does the paren mean in the rgl docs? [11:22:03] mkText Utt -> (Punct) -> (Text) -> Text Does she sleep? Yes. [11:24:16] it means it's optional [11:24:37] you can call mkText with Utt, Utt and Text, Utt and Punct or Utt, Punct and Text [11:29:48] tack :) [11:31:26] so I'm trying to extend the query language (ch 7) so that it can either just give a simple yes/no, or a yes [3 is prime]/no [4 is not prime] [11:32:54] i can obviously overload the yesno function [11:32:56] yesno = overload { [11:32:56] yesno : Utt -> Text ; [11:32:56] yesno utt = mkText (mkPhr utt) fullStopPunct ; [11:32:56] yesno : Utt -> Cl -> Text ; [11:32:56] yesno utt cl = mkText (mkPhr utt) fullStopPunct (mkText (mkPhr cl) fullStopPunct ) ; [11:33:18] } [11:33:18] but does this mean i have to include now two abstract functions for yes, and two for no [11:33:39] or I guess i could handle it in the linearization with a table/parameter setup [11:33:56] but i'm just wondering if there are any insights to what would be best practice [11:34:16] the best is to add new abstract functions [11:34:24] tack :) [11:34:46] only then do you have a chance to control when to output yes/no and when to output a more verbose explanation [11:34:56] so like this then [11:34:57] Yes : Answer ; [11:34:57] No : Answer ; [11:34:57] YesObj : Object -> Answer ; [11:34:57] NoObj : Object -> Answer ; [11:35:05] if you only wanted to accept both but linearise one, then you could make them as variants [11:35:18] but you seem to want to output both depending on the situation [11:35:29] indeed ! [11:35:45] I think you need to include the question, not just object [11:35:53] otherwise you don't know what to say [11:36:05] "yes, is prime" to a question whether it's even [11:36:19] or you can do even moe [11:36:29] YesIsPrime, YesIsEven, YesIsOdd : Object -> Answer [11:36:34] and same for NoIs... [11:36:52] no need to be afraid of adding lots of functions :-P [11:37:07] cool, i'll be experimenting all day then [11:37:17] i finally feel like i'm getting ahold of the RGL [11:37:45] it'll be simpler to add answers for each question as different abstract syntax funs, otherwise you'd need to retain a bunch of information in the lincat of Question [11:37:49] nice! [11:48:18] < Lycurgus> inariksit, funny that's just the kind of thing chomsky was on about in the mentioned talk: https://www.youtube.com/watch?v=X4F9NSVVVuw about 1:10 in [11:48:20] it's not really the same, chomsky's example and mine both have adverbials, but I'm talking about how one orthographical word contains elements from different constituents, and chomsky just seems to be talking about coordination [11:49:15] *** Quits: wmacmil (~wmacmil@c83-248-104-92.bredband.comhem.se) (Ping timeout: 265 seconds) [12:08:20] *** Joins: wmacmil (~wmacmil@84.216.157.27) [12:17:58] so one thing i'm confused about is when you say include a question [12:18:22] the question already has the "?" included in it's linearization [12:18:23] isNumericProp : Str -> NP -> Utt ; [12:18:23] isNumericProp even obj = mkUtt (mkQS (mkCl obj (mkAP (P.mkA even)))) ; [12:19:14] so it's not possible to do a YesIsProp : Question -> Answer ; [12:19:41] because the question has already been a kind of final form in its linearization [12:21:16] so the only solution is see is the YesIs* for each proposition on numbers, like you suggested above [12:47:23] *** Quits: wmacmil (~wmacmil@84.216.157.27) (Ping timeout: 260 seconds) [12:51:10] yes, if you wanted to do YesIsProp : Question -> Answer, then the lincat of Question must include, not just the ready string with "?" in the s field, but some more fields that can be used when building an Answer [12:51:30] but that's more complicated than just creating a new YesIs* answer for each question [14:35:44] *** Joins: wmacmil (~wmacmil@c83-252-138-144.bredband.comhem.se) [14:42:56] yes, if you wanted to do YesIsProp : Question -> Answer, then the lincat of Question must include, not just the ready string with "?" in the s field, but some more fields that can be used when building an Answer [14:42:58] but that's more complicated than just creating a new YesIs* answer for each question [16:02:18] ok, tack sa mycket @inari. [21:11:25] *** Quits: wmacmil (~wmacmil@c83-252-138-144.bredband.comhem.se) (Remote host closed the connection) [21:11:51] *** Joins: wmacmil (~wmacmil@c83-252-138-144.bredband.comhem.se)