[01:51:04] *** Quits: proteusguy (~proteusgu@cm-58-10-154-147.revip7.asianet.co.th) (Ping timeout: 272 seconds) [05:29:16] I had problems with the c-runtime finding libsg.so [05:29:27] I forgot how it started working. [05:30:08] I would try first to install without c-runtime. [05:30:58] *** Joins: proteusguy (~proteusgu@mx-ll-14.207.4-86.dynamic.3bb.co.th) [05:35:15] Are the directories in ld.so.conf, as done by ldconfig? [07:07:08] *** Quits: proteusguy (~proteusgu@mx-ll-14.207.4-86.dynamic.3bb.co.th) (Ping timeout: 245 seconds) [07:38:40] *** Joins: proteusguy (~proteusgu@mx-ll-14.207.4-86.dynamic.3bb.co.th) [10:43:02] drbean_: you need --extra-include-dirs=PATH and --extra-lib-dirs=PATH [10:43:11] if it doesn't find those [10:43:23] and of course, figure out the right path where they are installed :-P [10:44:15] and yeah anoe has done that already, so those errors seem to be something else! [10:45:04] I suggest you write (or search your error) on the mailing list https://groups.google.com/forum/#!forum/gf-dev [10:45:16] the author of the C runtime is there, but not on this channel [10:49:12] also, welcome to the channel and to GF! ^_^ as for deep learning "winning", depends what you consider the criteria; GF is suited for different kinds of problems [10:49:55] if you want to get published in top NLP conferences, doing GF isn't such a good way to get there [10:51:15] but if you want to model a small domain accurately and transparently, i.e. you see a bug and can fix it by changing (or adding) some rule(s), then GF is a good tool [12:15:16] *** Quits: stoopkid (uid137696@gateway/web/irccloud.com/x-uiotybabltahdgdz) (Quit: Connection closed for inactivity) [12:26:30] *** Quits: proteusguy (~proteusgu@mx-ll-14.207.4-86.dynamic.3bb.co.th) (Remote host closed the connection) [13:33:08] Yes. The error mentions pthread_mutex, That doesn't seem like something GF would be supplying. Or is it? [13:36:34] the error is about references to libpthread which is part of libc [13:56:08] *** Joins: proteusguy (~proteusgu@cm-58-10-154-147.revip7.asianet.co.th) [14:01:29] I am trying to work with gf-core/runtime/haskell, and I would like to import gf-rgl resources for a given language, do you have some documentation to do that [14:01:32] ? [14:01:43] (thanks btw for your reply) [14:11:52] (elsewhere than the book I mean) [14:15:11] anoe: you try to load the rgl from haskell? [14:15:17] yes [14:15:34] I am wondering how to load it [14:15:41] you first have to compile the grammar you want to a pgf file [14:16:06] and open this pgf with readPGF after [14:16:07] ? [14:16:21] exactly [14:16:29] ok, I see, thanks! [14:16:40] (I guess I see ^^) [14:31:39] you can compile a PGF like this [14:31:48] gf -make LangFoo.gf LangBar.gf ... [14:31:59] I was trying to use this: [14:32:08] gf -make --output-format=haskell QuestionEn.gf [14:32:14] gf -make --output-format=haskell QuestionEng.gf [14:32:25] that works too, then you also get Question.hs [14:32:40] but it answers: File Question.gf does not exist. [14:32:40] searched in: ../english/ [14:33:07] you need to either specify a path in the file, or set an environment variable GF_LIB_PATH [14:33:19] note that you need to do the whole LangEng.gf or at least GrammarEng.gf, with just Question you don't get much interesting stuff [14:33:41] e.g. you don't have any words in scope where you could apply the functions in QuestionEng [14:34:38] and not just words, e.g. QuestCl : Cl -> QCl ; -- does John walk needs a Cl as an argument, but in QuestionEng there are no functions that make a Cl [14:35:07] let me search an example of paths [14:35:13] thanks [14:35:55] --# -path=.:../abstract [14:36:08] in fact it's right here https://github.com/GrammaticalFramework/gf-rgl/blob/master/src/english/GrammarEng.gf [14:36:19] so if you just try to compile the whole LangEng [14:36:22] gf -make LangEng.gf [14:36:26] does that work? [14:37:04] yes [14:37:07] \o/ [14:37:51] are you using the haskell bindings? [14:37:52] gf -make --output-format=haskell LangEng.gf [14:37:54] works too :) [14:38:07] yeah, the output-format=haskell makes no difference to the produced pgf [14:38:44] reading the Lang.hs: very intersting! [14:38:49] did you get any problems with there being multiple modules named pgf, btw? [14:39:16] haha maybe I shjouldn't scare you with potential problems :-D but I remember at some point that being a problem, if one had installed the haskell bindings [14:39:41] not yet, will see and tell you if I encounter it [14:40:01] I think I will have to renmae the module Lang [14:40:05] into Eng.Lang [14:40:07] Fre.Lang [14:40:09] ... [14:40:14] no need [14:40:18] or Lang.Eng [14:40:21] you can have multiple concretes in a PGF [14:40:36] here's a code snippet I showed someone else on this channel a year ago https://gist.github.com/inariksit/4591fe1dabfad3fbf348846ce7c899ea [14:40:42] if I want to import in my haskell program ? [14:40:43] so ignore that the grammar is called differently [14:40:54] those are all different concrete syntaxes in the PGF [14:41:32] you can see a list of all concretes by using the function languages [14:42:04] ok [14:42:13] now there's only english [14:42:19] but so you can produce a multilingual pgf like this [14:42:29] gf -make english/LangEng.gf french/LangFre.gf … [14:42:34] (that was my next question, thx :) [14:43:13] File Prelude.gf does not exist. [14:43:13] searched in: french/ [14:43:34] add --# -path=.:../abstract:../prelude [14:43:40] to the top of LangFre.gf [14:43:50] or possibly GrammarFre.gf, I don't remember [14:44:10] or do you want to set up $GF_LIB_PATH now! :-D [14:44:28] Yes, I should I think [14:44:33] did you compile the rgl by using the makefile in the root directory? [14:44:46] I have used Setup.hs [14:44:49] okay [14:44:57] so you know where it installed everything? [14:45:10] yes [14:45:21] then just export $GF_LIB_PATH=/that/path [14:45:38] it should have folders alltenses, present and prelude in there [14:45:45] indeed [14:45:59] good [14:46:03] export GF_LIB_PATH=$(pwd) [14:46:04] works [14:46:08] yay [14:46:16] then you can also open gf shell anywhere and just do [14:46:23] > i alltenses/LangEng.gfo [14:46:34] or whatever language you want [14:48:10] great, thanks for help [14:48:19] (compiling right now) [14:48:22] no worries! just ask if you run in to more problems [14:51:03] how did you run into gf, btw? ^^ [14:51:38] I first developped an text-mining tool in Python/Javascript [14:51:52] Then I translated it into Haskell/Purescript [14:52:05] I discovered Gerard Huet work on Sanskrit [14:52:50] and then work of Aarne Ranta [14:53:57] nice! [14:54:37] yes, I am using pos-tagging tools (CoreNLP of Stanford) to extract NP [14:54:49] I am wondering if I could improve it with GF [14:56:15] you want to work with unrestricted text, I suppose? [14:56:29] yes [14:56:47] yeah, it might be difficult; in order to extract NPs, you need to parse the whole thing [14:57:15] sentence by sentence ? what do you mean with "the whole thing" ? [14:58:35] hah sorry that was unclear; I mean, if you have a sentence like "the jabberwocky froobled the napkin from the table", even if the GF lexicon covers napkin and table, and you could add some default rule that anything after "the" is likely a noun, it'd still need to parse this nonexistent verb [14:58:48] there is a chunking grammar in here https://github.com/GrammaticalFramework/wide-coverage , you could have a look at that [14:59:17] if you only need to find NPs and don't care that the rest of the sentence [15:01:34] (reading it) [15:02:20] okay : translate.probs [15:02:32] my blog posts can also be relevant, if you're planning on parsing string literals: https://inariksit.github.io/gf/2018/12/05/literals.html (background) and https://inariksit.github.io/gf/2019/01/26/literals-2.html (the actually helpful part) [15:03:22] translate.probs comes from the penn treebank [15:03:52] you can also compile a GF grammar without probabilities, it only affects parsing [15:03:58] i.e. in which order the trees are shown [15:10:33] do you have some example of how to use probas with GF ? [15:11:16] wait a sec [15:11:26] I mean hybrid system GF / Statistics ? [15:11:38] gf -make -probs=foo.probs [15:11:40] ah right [15:11:41] there are papers [15:11:46] another sec :-D [15:11:49] :) [15:12:38] https://www.aclweb.org/anthology/E14-1039 [15:16:15] (reading it) [15:20:08] sounds promising [15:21:54] (instead of the quality of the parsing - 8.4 ) [15:25:11] s/instead of/but/ [15:26:48] problem is hard indeed [15:27:56] yeah! [18:26:15] *** Quits: proteusguy (~proteusgu@cm-58-10-154-147.revip7.asianet.co.th) (Remote host closed the connection) [18:28:30] *** Joins: proteusguy (~proteusgu@cm-58-10-154-147.revip7.asianet.co.th) [21:13:27] *** Joins: stoopkid (uid137696@gateway/web/irccloud.com/x-gbtilukvepnkbwph)