[01:20:12] *** Quits: sirdancealot (~sirdancea@236.152.broadband3.iol.cz) (Ping timeout: 276 seconds)
[06:46:10] *** Joins: graphtheory (~textual@117.205.227.8)
[07:40:45] *** Quits: graphtheory (~textual@117.205.227.8) (Read error: Connection reset by peer)
[07:40:57] *** Joins: graphtheory (~textual@117.205.227.8)
[08:33:43] *** Quits: graphtheory (~textual@117.205.227.8) (Quit: My Mac has gone to sleep. ZZZzzz…)
[10:35:23] *** Joins: kook_ (~sirdancea@109.107.211.244)
[13:42:24] *** Joins: graphtheory (~textual@117.203.207.85)
[15:13:09] *** Joins: arademaker (user@nat/ibm/x-ibokelqtcgyvbssr)
[15:16:03] <arademaker> how can I test GF with a general purpose English Grammar ?
[15:30:03] *** Quits: arademaker (user@nat/ibm/x-ibokelqtcgyvbssr) (Ping timeout: 240 seconds)
[15:34:35] *** Joins: arademaker (user@nat/ibm/x-jlqforbhuecovdzw)
[15:35:02] <arademaker> Similar to HPSG online demo, http://erg.delph-in.net/logon, I would like to test GF with a sample English sentence
[15:50:55] *** Quits: graphtheory (~textual@117.203.207.85) (Quit: My Mac has gone to sleep. ZZZzzz…)
[15:54:03] *** Quits: arademaker (user@nat/ibm/x-jlqforbhuecovdzw) (Ping timeout: 250 seconds)
[16:23:48] *** Quits: kook_ (~sirdancea@109.107.211.244) (Ping timeout: 250 seconds)
[16:32:01] *** Joins: graphtheory (~textual@117.203.207.85)
[16:43:18] <inariksit> if arademaker comes back, link them this: http://cloud.grammaticalframework.org/wc.html
[16:43:37] <inariksit> daherb: why isn't your bot here now that we need her! :-P
[16:45:19] *** Quits: drbean (~drbean@124.219.82.41) (Ping timeout: 244 seconds)
[16:52:24] *** Joins: drbean (~drbean@124.219.82.41)
[16:56:43] <daherb> inariksit: sorry, it crashed again and i now try to fix it. and now i am in the cabal dependency hell for the moment
[16:57:08] <inariksit> hehe
[16:57:29] <inariksit> yeah no worries, I'm just complaining and you're the one who's actually maintaining
[16:58:05] <inariksit> how's your SLTC paper going?
[17:00:01] *** Joins: annaaerinteenbot (~annaaerin@cse-principia.cse.chalmers.se)
[17:00:20] <inariksit> woo
[17:00:36] <daherb> i just hope it doesn't crash again
[17:00:36] <inariksit> annaaerinteenbot: tell arademaker here you can try out sentences http://cloud.grammaticalframework.org/wc.html
[17:00:37] <annaaerinteenbot> inariksit: i will tell arademaker
[17:13:09] <inariksit> daherb: I actually thought I could try to write something about my GF stuffs for SLTC
[17:13:17] <inariksit> I just started :-D
[17:13:46] <inariksit> it'll be a nice exercise in writing bullshit in 3 days ^^
[17:13:50] <daherb> about hungarian?
[17:13:52] <inariksit> yeah
[17:15:35] <daherb> my paper is basically down to 3 pages but i'd like to improve it further because it is not very coherent yet
[17:16:34] <inariksit> I can read it if you want
[17:17:11] <daherb> maybe later tomorrow
[17:20:23] *** Joins: arademaker (user@nat/ibm/x-hncbbjlusennhaix)
[17:20:23] <annaaerinteenbot> arademaker: inariksit wants me to tell you  here you can try out sentences http://cloud grammaticalframework org/wc html
[17:20:23] <annaaerinteenbot> inariksit: transmitted your message to arademaker
[17:20:37] <inariksit> arademaker: here's a working link http://cloud.grammaticalframework.org/wc.html
[17:20:53] <inariksit> it's a translator based on the Resource Grammar Library, and some extensions
[17:22:39] <daherb> inariksit: i should probably be more careful with the stripping of punctuation marks
[17:22:47] <arademaker> thanks.
[17:23:11] <arademaker> any reference about the use o GF for semantic parsing + information extraction?
[17:23:33] <inariksit> hmm
[17:23:44] <arademaker> I am interested in extract meaning representations from text.
[17:23:49] <inariksit> I'm sure there is, but I'm the wrong person to ask
[17:23:55] <inariksit> I just like to write resource grammars :-P
[17:25:50] <arademaker> ;-)
[17:26:18] <arademaker> So you would be a nice person to ask direction about how to contribute to develop Portuguese GF grammar.
[17:27:07] <inariksit> yes!
[17:27:30] <inariksit> does this count as meaning from text? :-D it checks that you're saying right things about a pizza https://github.com/inariksit/esslli2015-monads/tree/master/pizza-semantics
[17:28:02] <inariksit> about portuguese, we had someone in the summer school 2015 who was doing portuguese, but nobody's heard of him since
[17:29:45] <arademaker> Consider a sentence "The depths of the wells in the Yates field vary from 900 to 1700 ft."
[17:30:10] <arademaker> I would like to extract fact such as "Yates is a Field" and "The depth of the wells in Yates is between 274.32 to 518.16 meters"
[17:31:43] <inariksit> for the first, you can add a rule that whenever you parse something like
[17:31:44] <inariksit> (DetCN (DetQuant DefArt NumSg) (PossNP (UseN field_2_N) (UsePN (SymbPN (MkSymb "Yates"))))))
[17:32:33] *** Quits: annaaerinteenbot (~annaaerin@cse-principia.cse.chalmers.se) (Ping timeout: 240 seconds)
[17:32:38] <inariksit> then you say "the two arguments of the PossNP have this kind of relation with each other"
[17:32:57] <inariksit> ok this is going to be a crappy rule, also no idea why id parsed it as "the field of Yates"
[17:33:29] *** Joins: kook_ (~sirdancea@236.152.broadband3.iol.cz)
[17:33:39] <arademaker> the output in the demo using Eng->Eng is http://paste.lisp.org/display/325342
[17:34:09] <inariksit> ah yeah that's just fallback to chunks, it's not nice to extract any meaning from that
[17:34:56] <inariksit> if you have a limited domain, then it's definitely worthwhile to write your own controlled language, and make the rules very precise
[17:35:06] <arademaker> so I imagine that a more 'semantic' oriented structure is necessary for meaning extraction, right?
[17:35:11] <inariksit> yeah
[17:35:23] <inariksit> if you get the yellow parse (try just "the wells in the Yates field are big")
[17:35:35] <inariksit> that's also quite fine, you have the whole sentence parsed syntacticall
[17:35:36] <inariksit> y
[17:35:57] <inariksit> and you can be somewhat confident that the things are related to each other properly
[17:36:15] <inariksit> with chunks, it's just [the wells] [in the yates field] [...]
[17:36:34] <inariksit> in the yellow parse, you have an actual tree, not a flat list of chunks
[17:36:37] <arademaker> got it. So red means incomplete.
[17:36:39] <arademaker> ?
[17:36:52] <inariksit> you could write your own grammar and it would show as green
[17:37:00] <inariksit> yeah, red means that GF grammar couldn't parse the whole thing
[17:37:07] <arademaker> orange actually.
[17:38:23] <inariksit> hehe :-P
[17:38:34] <arademaker> For processing technical english from Oil&Gas, it would be possible to expect some results expanding the lexical of the already available English grammar?
[17:39:15] <inariksit> at least for getting chunk parses, that would definitely help
[17:39:57] <arademaker> I see. About the example http://paste.lisp.org/display/325342, why 900 is parsed different than 1700 ?
[17:39:58] <inariksit> oh wow that's stupid, I changed the sentence to "The depth of the wells in Yates is between 274 to 518 meters" and then it gets a yellow parse
[17:40:17] <inariksit> oops sorry I tried wrong thing
[17:40:55] <inariksit> about 1700 vs 900, haha I see; so all the numbers can be parsed with the number grammar (just glue digits together), but also as symbols
[17:41:18] <inariksit> actually, anything can be parsed as symbols, so sometimes you get really ridiculous parses where the word "the" is a symbol >_>
[17:41:28] <inariksit> in that particular case, no idea
[17:42:58] <inariksit> oh wow also, this works: "The depths of the wells in the Yates field vary from 900 to 900 feet"
[17:43:00] <arademaker> hum.. But it is traceable, right?
[17:43:04] <inariksit> so why doesn't 1700 work O_o
[17:43:30] <inariksit> 999 works, 1000 doesn't
[17:44:20] <inariksit> I should be going home
[17:44:36] <inariksit> but if you have questions, go ahead
[17:44:43] <inariksit> I can stay for some time to help
[17:44:46] <arademaker> thanks for the chat. I would enjoy have more references whenever possible!
[17:45:13] <arademaker> no, please go ahead. Go home, I am reading some texts. We can talk more other day.
[17:45:29] <inariksit> cool :) hope you find something!
[17:45:57] <inariksit> did you search from the google group too?
[17:52:55] <inariksit> https://groups.google.com/forum/#!topic/gf-dev/OWurmr3MaVA I don't know if this is what you're after
[17:53:57] <inariksit> (also it's from 2008 and nobody in the thread gets it to compile :-P)
[17:56:03] <arademaker> I will take a look, thanks
[19:06:12] *** Quits: graphtheory (~textual@117.203.207.85) (Ping timeout: 276 seconds)
[19:35:47] <inariksit> arademaker: also you could take a look at chapter 8.8 in the GF book
[19:37:10] <inariksit> for translating the syntactic structures into semantics
[19:37:54] <inariksit> it looks very strange, and probably in your use case, you would rather use the PGF library from an external programming language
[19:38:33] <inariksit> but the point where a sentence like "every woman walks" gets translated into a logical semantics thing
[20:00:28] <inariksit> sorry about the flood, I'm learning things myself as well ^^ this may also be interesting: http://i12www.iti.uni-karlsruhe.de/~key/oclnl/lacl05.pdf
[20:00:43] <inariksit> I suggest you check page 6 for Figure 3, to see the example generated text
[20:01:07] <inariksit> and starting from page 7, chapter 4 System overview
[20:03:21] <inariksit> it's the opposite to your use case; they start from a specification language and generate natural language, but if you've got a GF grammar, everything can be the source or the target
[20:03:56] <inariksit> also they mention using an external parser and generating GF trees, and also manipulating the trees in order to create more natural sounding text
[20:05:24] *** Joins: arademak` (user@nat/ibm/x-olomaggeengitcsr)
[20:05:57] <arademak`> @inariksit thanks. Interesting article.
[20:08:25] *** Quits: arademaker (user@nat/ibm/x-hncbbjlusennhaix) (Ping timeout: 250 seconds)
[20:09:05] *** Joins: arademaker (user@nat/ibm/x-hzdecpoucxacfomq)
[20:10:02] <inariksit> arademaker: is it easy to follow, being a beginner in GF? :)
[20:11:56] <arademaker> yes
[20:13:19] <arademak`> beginner in GF but not completely new to related areas!
[20:13:52] *** Parts: arademaker (user@nat/ibm/x-hzdecpoucxacfomq) ()
[20:14:29] <inariksit> ok yeah that makes it easier surely ^^ what's your background?
[20:15:26] <arademak`> http://arademaker.github.io/publications.html ! ;-) Logic, Proof Theory, We maintain the Portuguese Wordnet part of the GWA..
[20:16:07] <arademak`> After my phd I started work with NLP, mostly interested in semantic parsing and information/knowledge extraction.
[20:16:36] *** Quits: arademak` (user@nat/ibm/x-olomaggeengitcsr) (Quit: ERC (IRC client for Emacs 24.5.1))
[20:16:42] *** Joins: arademaker (user@nat/ibm/x-tvhqtparobjfpsmn)
[20:17:46] <inariksit> cool :) hehe I should've googled you first, I only remembered that you'd posted on the GF list and said you only recently started learning GF
[20:19:03] <arademaker> no problem! ;-) I am still trying to identify you from the nick name
[20:19:17] <inariksit> http://www.cse.chalmers.se/~inari/
[20:19:20] <inariksit> that's me!
[20:19:54] <arademaker> thanks!
[20:20:21] <arademaker> grammars + SAT, cool! I liked it.
[20:21:00] <arademaker> I would be very interesting in connecting the subhect with provers linke Coq and Lean
[20:22:28] <arademaker> s/subhect/subject
[20:25:23] <inariksit> I don't know anyone who's been doing that recently
[20:25:33] <inariksit> but again I may just have missed someone
[20:44:16] <inariksit> this may be interesting too https://www.aclweb.org/anthology/W/W14/W14-1401.pdf
[20:59:47] *** Joins: annaaerinteenbot (~annaaerin@cse-principia.cse.chalmers.se)
[21:10:21] <inariksit> annaaerinteenbot: hej!
[21:10:21] <annaaerinteenbot> what do you want to accomplish by saying: "hej"?
[21:12:05] *** Quits: arademaker (user@nat/ibm/x-tvhqtparobjfpsmn) (Ping timeout: 244 seconds)