[04:41:20] *** Quits: JuanDaugherty (~juan@98.4.124.117) (Quit: Ex Chat) [05:03:04] *** Quits: stoopkid__ (uid137696@gateway/web/irccloud.com/x-prfychaifrxqfupu) (Quit: Connection closed for inactivity) [05:08:23] mbrock inariksit: I've been waiting for a response for that too :) [05:10:03] inariksit: nobody_NP is used with double negation in Turkish [05:10:35] in this case, what would `(PredVP (UsePron he_Pron) (UseV sing_V))` linearize to? [05:11:27] should I make it so that it linearizes to “nobody sings” (which is not the correct things) and then handle then negation in the implementation of UttS? [05:11:56] Or should it linearize to the correct thing i.e., “nobody doesn't sing” [05:15:25] inariksit: oh I just saw your message about the testing thing, it sounds really cool! [05:19:41] so this is for generating examples from a core subset of the RG abstract and checking if those are correct or not, did I understand that correctly? [05:23:02] how are the expected outputs given? [05:28:01] I started doing testing like this: https://github.com/ayberkt/GF/blob/testing-tur/testsuite/libraries/test_tur.sh [05:43:45] I am cloning it now, is it working for any grammar? [05:46:00] this was exactly, what I was talking about: https://github.com/inariksit/GF-testing/blob/master/src/Lib.hs#L15 ! [05:48:35] is the intended workflow generating tests though or manually specifying cases and outputs? if it is the latter, there is nothing preventing us from using the machinery here for the latter too right? [05:50:32] For Haskell, I usually use hspec for both QuickCheck and unit tests and I really like it; it would be like that [05:51:25] also, can someone suggest me a notification system for IRC? I keep forgetting to check because I haven't set up some kind of notification thing. [05:52:25] I have an SMTP server running on my server so I can mail myself, I just need to figure out how to run commands when my name is mentioned... [09:15:12] I'm lazy these days and pay 4 EUR / month for irccloud.com which has mobile apps with push notifications [09:55:24] ayberkt: yeah it looks like the same workflow, the difference is just that the method of generating the example trees should be Fancy and Automatic and stuff like that ^^ [09:56:13] (but yeah as you might notice, I'm just using this function https://github.com/inariksit/GF-testing/blob/master/src/Lib.hs#L116 which takes just the first tree of a given type) [09:56:28] we're working on some low-level PGF hacking [09:57:27] so, not manually specifying cases. for outputs it depends; I'm using mine now just to check for differences [09:57:50] i.e. before I changed some function, the linearization was X; now I change the function and linearize everything again and diff to the original X [09:57:57] but it may have been completely wrong in the original too [09:58:39] as for nobody_NP, it's just one of the things where the RGL is imperfect, and an application grammarian should write their own thing [09:59:59] the negation on the sentence level depends on the argument of type Pol: https://github.com/GrammaticalFramework/GF/blob/master/lib/src/abstract/Sentence.gf#L59-L62 [10:00:13] and so it should be possible to generate all the combinations of subjects and polarities [10:00:21] so "nobody sleeps / nobody doesn't sleep" [10:00:58] it's a well known thing that RGL doesn't produce idiomatic translations, and often overgenerates and is ridiculous, and hopefully less often is outright wrong [10:01:13] *** Joins: kindaro (bca20e9d@gateway/web/freenode/ip.188.162.14.157) [10:01:21] *** Quits: kindaro (bca20e9d@gateway/web/freenode/ip.188.162.14.157) (Client Quit) [10:01:24] so if it's completely ungrammatical to say "nobody sleeps", then it's just one of those things where it's wrong [10:01:51] if it's just strange, then that's just the everyday RGL experience :-P [10:02:35] but you shouldn't try to make a humble nobody_NP try to affect things on the sentence level negation [10:03:11] I have another example: here in Structural there's a possess_Prep https://github.com/GrammaticalFramework/GF/blob/master/lib/src/abstract/Structural.gf#L72 [10:03:37] in basque, I wanted to first make the possess_Prep different according to whether it combines with an animate or inanimate noun [10:04:25] but that made things slower and awkward; and then I noticed that there is this function PossNP in Noun: https://github.com/GrammaticalFramework/GF/blob/master/lib/src/abstract/Noun.gf#L146 [10:05:06] so I implemented the test there: if the NP is animate, then use one string, if it's inanimate, then use another [10:05:47] okay this was not really related, but just as an example that sometimes it's just okay to leave things overgenerating or underspecified or strange, but there can be ways to implement the same thing in a nicer way [10:06:14] for IRC notifications, I don't mind you answering with a delay :-P if I had something really urgent, I'd just email you! [10:06:24] I don't have any notification thing, I just am here all the time :-D [10:07:01] I'm using irssi, and on my mobile I use irssi connectbot, it's just a routine that if I'm on a tram or something, I check IRC [10:07:12] and if I'm on a computer, I just have it open and check every now and then [10:08:54] 04:43:43 < ayberkt> I am cloning it now, is it working for any grammar? <- try and change the path /Users/inari/... into wherever you have your GF source code, and try ./testLang (basque|dutch) , then you'll see ^^ [10:09:06] I wouldn't be surprised if it still doesn't work :-P [10:12:08] (sorry, ./testLang.sh ) [11:10:03] *** Joins: JuanDaugherty (~juan@98.4.124.117) [11:20:48] *** Joins: proteusguy (~proteus-g@2405:9800:bc10:1ca:acb1:27dd:1875:4bcc) [11:25:56] *** Quits: proteusguy (~proteus-g@2405:9800:bc10:1ca:acb1:27dd:1875:4bcc) (Ping timeout: 246 seconds) [12:08:47] *** Joins: proteusguy (~proteus-g@180.183.48.154) [12:55:05] *** Quits: proteusguy (~proteus-g@180.183.48.154) (Ping timeout: 240 seconds) [13:07:17] *** Joins: proteus-guy (~proteusgu@180.183.10.249) [19:24:49] *** Joins: stoopkid__ (uid137696@gateway/web/irccloud.com/x-jjjhuoezxqtqcpto) [19:43:31] *** Quits: proteus-guy (~proteusgu@180.183.10.249) (Remote host closed the connection) [20:07:10] *** Joins: proteusguy (~proteus-g@2001:44c8:4485:1983:acb1:27dd:1875:4bcc) [20:20:02] *** Quits: proteusguy (~proteus-g@2001:44c8:4485:1983:acb1:27dd:1875:4bcc) (Ping timeout: 246 seconds) [20:42:00] *** Quits: koom (~sirdancea@109.107.211.244) (Ping timeout: 246 seconds) [21:31:38] *** Joins: proteusguy (~proteus-g@2405:9800:bc10:1ca:acb1:27dd:1875:4bcc)