[00:26:04] <mmehdig> @inariksit, now I understand that "tire_V" was not a good example. better example is the simplest verb like "love_V" (or "hate_V"). They need more than "awkward translation". I believe I can write a set of new rules for Compound Verbs based on this theory: http://persian.nmelrc.org/pvc/CompoundVerbs_DabirMoghaddam.pdf
[00:34:16] <spectie> don't you have all those light verb constructions in persian too
[00:34:20] <spectie> like 'X do'
[00:37:18] *** Quits: drbean (~drbean@sac.nuu.edu.tw) (Ping timeout: 252 seconds)
[00:37:49] *** Joins: drbean (~drbean@113.196.171.151)
[00:44:12] <mmehdig> well "X do" is just one form of all these compound verbs
[00:44:22] *** Joins: Eidel_ (~eidel@c83-249-247-253.bredband.comhem.se)
[00:45:29] <mmehdig> I want to follow this list: http://persian.nmelrc.org/pvc/compounds.php?sort=dabir&sem=0&theme=0
[00:45:56] <spectie> yeah
[00:47:02] <mmehdig> 13 group, but it's not hard because we already have mkVerb in GF which works fine ...
[00:47:14] *** Quits: Eidel (~eidel@c83-249-247-253.bredband.comhem.se) (Ping timeout: 240 seconds)
[00:51:00] <mmehdig> we don't proper opers for these formats in resource grammar: (ADJ + aux budan), (ADJ + aux shodan), (Past participle + passive aux)
[00:55:27] <mmehdig> they are important in lexicon. more importantly, we found out that a set of these verbs needs which only has Prepositional Objects needed immediate attention.
[01:00:31] <mmehdig> apparently current implementation didn't consider compound verbs without "ra" (the accusative marker). this "ra" grammar is specific for Persian there is nothing similar in turkish
[01:02:11] <spectie> really?
[01:02:24] <spectie> there is accusative in turkic
[01:02:51] <mmehdig> but there is no postpositional case marker like "ra"
[01:03:03] <spectie> usually -nI
[01:03:18] <spectie> you can consider it a case or a postposition
[01:03:21] <spectie> according to taste ;)
[01:04:07] <spectie> i can't remember if -ra ending in persian is optional with transitive verbs
[01:04:11] <spectie> like -nI in turkic
[01:05:15] <mmehdig> "ra" is not like a clitic, but it has alternatives like "-o", "-ro"
[01:06:23] <mmehdig> well some verbs like "hate_V" they don't accept "ra"
[01:07:12] <spectie> they take a PP ?
[01:07:35] <mmehdig> "I hate something" in persian has translation of "I am in hate with something"
[01:07:40] <spectie> aye
[01:07:45] <spectie> but that is not transitive
[01:07:46] <mmehdig> yeah
[01:07:51] <spectie> that's an entirely different construction :)
[01:08:19] <mmehdig> ok but the construction of "hate_V" is like this
[01:08:58] <mmehdig> (hate_N + toBe)
[01:09:40] <mmehdig> man az aan motenafer-am
[01:09:52] <spectie> az = from ?
[01:09:58] <mmehdig> yes
[01:10:09] <mmehdig> it means "I hate that"
[01:10:56] <mmehdig> or as I said about long form of "to be": man az aan motenafer hastam
[01:13:15] <mmehdig> and more about the "ra", there is a case that if you drop it in sentence like: "I saw the book" will become "I saw a book"
[01:13:40] <mmehdig> "man ketab ra didam" = "I saw the book"
[01:13:49] <mmehdig> "man ketab didam" = "I saw a book"
[01:14:32] <spectie> yeah
[01:14:35] <spectie> that's how it works in turkic too
[01:14:51] <spectie> the overt accusative is somehow more definite than the non-overt one
[01:15:14] <spectie> :)
[01:16:56] <mmehdig> maybe I should ignore this structure without "ra", because it sounds unusual.
[01:17:46] <spectie> more unusual than "all surfers eat a tasty coconut"
[01:17:47] <spectie> ?
[01:17:51] <spectie> ;)
[01:17:58] <spectie> brb
[01:19:27] <spectie> back
[01:19:33] <spectie> ok nn!
[01:20:04] <mmehdig> well in some articles that I read today, dropping the "ra", means connecting the noun to the verb, which is a way to generative new compound.
[01:20:43] <mmehdig>  "ketab didam" can become a book-seeing activity
[01:21:40] <mmehdig> to generate*
[01:23:00] <mmehdig> here it is the article: http://www.ensani.ir/storage/Files/20101125162018-146.pdf
[01:24:22] <mmehdig> the other set of articles are in persian.
[01:24:58] <mmehdig> this one actually is focused on a set of well-known verbs
[03:29:12] *** Quits: drbean (~drbean@113.196.171.151) (Ping timeout: 276 seconds)
[03:34:11] *** Joins: drbean (~drbean@sac.nuu.edu.tw)
[05:13:41] *** Quits: mmehdig (~mmehdig@cust-95-80-47-120.csbnet.se) (Remote host closed the connection)
[08:51:45] *** Joins: mmehdig (~mmehdig@cust-95-80-47-120.csbnet.se)
[08:59:57] <mmehdig> I have a question in english grammar.
[09:01:39] <mmehdig> "he sends flowers to him": NP -> V3 -> NP -> NP -> Cl
[09:02:16] <mmehdig> now what about "he sends him flowers"
[09:04:39] *** Quits: spectie (~fran@unaffiliated/spectie) (Ping timeout: 258 seconds)
[09:14:57] *** Joins: spectie (~fran@unaffiliated/spectie)
[09:19:38] <spectie> mmehdig, yep?
[09:20:00] <mmehdig> I think I found my answer! :D in the sentence "he sends her flowers", "sends" should interpret as "send_V2" instead of "send_V3"
[09:20:26] <spectie> what are V2 and V3 ?
[09:21:11] <Masx> v2 is transitive, v3 ditransitive
[09:21:16] <mmehdig> yes
[09:21:19] <Masx> but I think both sentences are V3?
[09:22:32] <mmehdig> when I try it on wide coverage translation, it interpret it as send_V2
[09:22:58] <mmehdig> but we don't have send_V2 in english resource grammar
[09:23:07] <spectie> as in "he sends [SN [Det her] [N flowers]"
[09:23:11] <spectie> ..to the moon
[09:24:14] <mmehdig> it's how CompSlash  works
[09:24:59] <mmehdig> or at least Slash2V3 and Slash3V3 force you to have PP for V3 verbs
[09:27:44] <mmehdig> so ...  do you think it's correct to say it's send_V2 with an optional complement?
[09:30:22] <mmehdig> "someone gave me flowers"
[09:32:52] <mmehdig> now this sentence works fine in translator project but it doesn't work on main project.
[09:35:14] <mmehdig> it was my bad for previous sentence, "her flowers" was an ambiguous phrase.
[09:37:21] <mmehdig> then the question is there. why can't we have "he gave me flowers"?
[09:38:35] <mmehdig> an translator marks it yellow
[09:38:38] <mmehdig> and*
[09:45:50] <mmehdig> so ... is it ungrammatical to say "he gave me a flower" ?
[09:47:29] <mmehdig> apparently this structure is not supported in English resource grammar.
[10:46:08] <spectie> that is grammatical
[10:46:09] <spectie> brb
[11:11:49] <mmehdig> I think I need to understand these kinds of sentences in GF in order fix the problem in Persian.
[11:13:39] <Masx> mail aarne / ping inariksit ? :)
[12:43:05] *** Joins: chru (~christina@2001:638:504:2099:d148:8105:9a85:129a)
[12:50:26] <inariksit> 01:19:28 < mmehdig> maybe I should ignore this structure without "ra", because it sounds unusual.
[12:50:29] <inariksit> 01:20:19 < spectie> more unusual than "all surfers eat a tasty coconut"
[12:50:32] <inariksit> :---D
[12:50:51] <inariksit> hi everyone! lots of backlog
[12:50:52] <inariksit> ^^
[12:51:51] <inariksit> mmehdig: I think the wide coverage translator uses the penn treebank probabilities, and they might be sound or might not
[12:52:21] <inariksit> and if V2s are favoured over V3s, then it just gets higher probability to parse send_V2 + PP
[12:52:46] <mmehdig> yes I figured out that sample
[12:52:57] <inariksit> about the yellow colour, it means that the translation comes from the RGL, not from a domain-specific grammar
[12:53:10] <mmehdig> but in sentences like: "he sent me an email"
[12:53:48] <mmehdig> why there is no grammar for sentences like "he sent me an email" ?
[12:54:22] <mmehdig> GF only accepts "he sent an email to me"
[12:54:38] <inariksit> yeah, it's defined in lexicon as  send_V3 = dirV3 (irregV "send" "sent" "sent") toP ;
[12:55:06] <inariksit> the RGL is supposed to be kept without variants
[12:55:09] <inariksit> or at least the core
[12:55:31] <inariksit> in a big lexicon send_V3 could have both variants, with and without "to"
[12:56:00] <inariksit> you're completely free to define your own verb with two direct objects
[12:57:36] <inariksit> this just stems from the fact that RGL wasn't originally supposed to be used for wide-coverage translation, just a simple and predictable resource, if you want a NP with indefinite article you write "mkNP a_Det x" and if you want without you use MassNP instead, and it's the problem of an application grammar writer to decide when and in which language use which
[12:58:15] <inariksit> or if you want to use "send" with two direct objects or one with "to"
[12:59:17] <mmehdig> then "send_V3" should have two variants right?
[12:59:56] <mmehdig> something like: send_V3 = dirV3 (irregV "send" "sent" "sent") toP | dirV3_2DO (irregV "send" "sent" "sent") toP
[13:00:20] <mmehdig> I don't know how does dirV3 works
[13:01:04] <inariksit> yeah that's just a low-level function, mkV3 is overloaded with some options
[13:01:05] <inariksit> http://www.grammaticalframework.org/lib/doc/synopsis.html#toc90
[13:01:24] <inariksit> so actually just mkV3 (irregV "send" "sent" "sent")  will do the trick
[13:01:28] <inariksit> this toP just means to_Prep
[13:03:08] <inariksit> so yeah, after modifying that in LexiconEng, it parses
[13:03:09] <inariksit> Lang> p -cat=Cl "she sent him flowers"
[13:03:09] <inariksit> PredVP (UsePron she_Pron) (ComplSlash (Slash2V3 send_V3 (UsePron he_Pron)) (DetCN (DetQuant IndefArt NumPl) (UseN flower_N)))
[13:03:12] <inariksit> PredVP (UsePron she_Pron) (ComplSlash (Slash3V3 send_V3 (DetCN (DetQuant IndefArt NumPl) (UseN flower_N))) (UsePron he_Pron))
[13:03:55] *** Joins: jmvanel (~jmvanel@37.161.73.130)
[13:04:03] <mmehdig> yes it worked here too! :D
[13:04:20] <inariksit> \o/
[13:04:58] <mmehdig> no wait! what are these?
[13:05:58] <mmehdig> linearization didn't work!
[13:07:18] <mmehdig> "she sent him to flowers" :))
[13:09:09] <mmehdig> now we need to figure out the problem in Slash3V3 or Slash2V3 right?
[13:09:56] <inariksit> are you using variants?
[13:10:01] <mmehdig> yes
[13:10:08] <inariksit> yeah, they appear to be buggy
[13:10:19] <inariksit> like chru's "all surfers eats" or "every surfer eat"
[13:10:38] <inariksit> I just replaced mine with a single   mkV3 send_V
[13:10:57] <inariksit> Lang> p  "she sent him flowers" | l
[13:10:58] <inariksit> she sent him flowers
[13:11:59] <inariksit> I need to do other stuff now, I might be here in the evening 21->, and from tomorrow on I'll be travelling
[13:12:12] <mmehdig> but we need both variants at the end, right? "she sent flowers to him"
[13:12:22] <mmehdig> ok ok. thank you :)
[13:12:29] <inariksit> in RGL the policy is to avoid variants; you can ask aarne more about that
[13:12:38] <inariksit> good luck!
[13:12:44] <mmehdig> ok thank you :)
[13:48:28] *** Quits: chru (~christina@2001:638:504:2099:d148:8105:9a85:129a) (Quit: Leaving.)
[13:48:47] *** Joins: chru (~christina@elpollodiablo.TechFak.Uni-Bielefeld.DE)
[14:09:01] *** Quits: jmvanel (~jmvanel@37.161.73.130) (Ping timeout: 258 seconds)
[15:00:59] *** Joins: jmvanel (~jmvanel@37.161.73.130)
[15:34:36] *** Quits: jmvanel (~jmvanel@37.161.73.130) (Ping timeout: 276 seconds)
[16:18:33] *** Joins: myoriel (~Myoriel@aftr-88-217-180-130.dynamic.mnet-online.de)
[17:41:39] *** Quits: mmehdig (~mmehdig@cust-95-80-47-120.csbnet.se) (Remote host closed the connection)
[18:13:58] *** Joins: jmvanel (~jmvanel@229.106.115.78.rev.sfr.net)
[18:35:19] *** Joins: mmehdig (~mmehdig@cust-95-80-47-120.csbnet.se)
[18:37:35] *** Quits: mmehdig (~mmehdig@cust-95-80-47-120.csbnet.se) (Remote host closed the connection)
[19:06:33] *** Parts: chru (~christina@elpollodiablo.TechFak.Uni-Bielefeld.DE) ()
[21:41:35] *** Quits: jmvanel (~jmvanel@229.106.115.78.rev.sfr.net) (Quit: Quitte)
[22:05:07] *** Joins: jmvanel (~jmvanel@229.106.115.78.rev.sfr.net)
[22:38:05] *** Joins: mmehdig (~mmehdig@cust-95-80-47-120.csbnet.se)
[22:42:33] *** Quits: mmehdig (~mmehdig@cust-95-80-47-120.csbnet.se) (Ping timeout: 252 seconds)
[22:51:20] *** Joins: mmehdig (~mmehdig@cust-95-80-47-120.csbnet.se)
[23:28:41] *** Quits: myoriel (~Myoriel@aftr-88-217-180-130.dynamic.mnet-online.de) (Ping timeout: 258 seconds)