[12:18:02] *** Joins: vinit-ivar (~vinit@122.169.11.109) [12:19:34] *** Quits: vin-ivar (~vinit@182.70.10.39) (Ping timeout: 244 seconds) [12:41:06] *** Quits: spectre (~fran@115.114.202.84.customer.cdi.no) (Ping timeout: 276 seconds) [13:03:57] *** Joins: spectre (~fran@dhcp674-ans.wifi.uit.no) [13:13:46] *** Quits: spectre (~fran@dhcp674-ans.wifi.uit.no) (Read error: Connection reset by peer) [14:45:18] *** Joins: spectre (~fran@dhcp800-ans.wifi.uit.no) [15:19:17] *** vinit-ivar is now known as vin-ivar [16:24:04] *** Joins: worldsayshi (~per.frede@h-5-150-197-146.na.cust.bahnhof.se) [16:42:24] *** Quits: spectre (~fran@dhcp800-ans.wifi.uit.no) (Ping timeout: 250 seconds) [16:45:38] *** Joins: spectre (~fran@dhcp800-ans.wifi.uit.no) [16:50:55] inariksit, spectre - what sort of stuff do you folks think I ought to implement before the 1st? I think I should come up with a decentish roadmap, of sorts [16:51:29] i want to improve quickbuilding corpora, first [17:06:25] *** Quits: worldsayshi (~per.frede@h-5-150-197-146.na.cust.bahnhof.se) (Quit: worldsayshi) [17:09:39] *** Quits: spectre (~fran@dhcp800-ans.wifi.uit.no) (Ping timeout: 265 seconds) [17:09:54] vin-ivar: hmm [17:11:35] is it easier to work in GF-apertium direction in general? [17:11:42] or just for verb/noun subcategories? [17:11:54] like vblex <> V V2 V3 VS VV etc. [17:14:58] it's easy for adjectives, because they really don't do much except contextually decline [17:15:04] and should be easy for adverbs too [17:15:21] GF has Adv, AdV and AdA [17:15:52] ^i don't know if I mentioned it here, I managed to scrape around 2500 nouns from apertium-dan and add them to GF [17:16:10] cool! [17:17:50] AdA = adadjective? [17:19:39] hm, i suppose making a manual tag mapping would make sense [17:19:45] spectre reckons there's no need to learn it [17:22:12] yeah, I would imagine so too, the tagsets are not so huge [17:26:30] yeah [17:26:31] can you start with some simple text that can be parsed in GF, and just start by doing some simple conversion, like Masc -> , Pres -> and see how far you get with that= [17:26:45] okay, i'll do that for spanish [17:26:46] for instance the story, if you like :-P [17:29:59] *** Quits: vin-ivar (~vinit@122.169.11.109) (Read error: No route to host) [17:30:54] *** Joins: vin-ivar (~vinit@122.169.11.109) [17:31:08] sorry [17:31:12] laptop decided to give up [17:31:14] on life [17:31:28] anyway, what I was thinking of was modifying the output of the morph analysis to match apertium's tags [17:31:50] is that what you said? I think i missed a couple of messages :P [17:35:12] *** Joins: spectre (~fran@115.114.202.84.customer.cdi.no) [17:35:39] vin-ivar: yep :) [17:36:02] so just try a simple baseline and run it on some texts, if that works, great, try with more complicated texts [17:36:29] also if all works fine with full GF parses, try with chunking; does it still work if the GF side doesn't parse the text completely [17:36:36] sounds cool [17:36:50] that reminds me [17:37:22] yeah? [17:37:30] would you happen to have a Spanish pgf file with you? my laptop sometimes dies when I try to make one [17:38:05] of TranslateSpa or just LangSpa? [17:38:07] I can make one [17:38:28] TranslateSpa, LangSpa shouldn't be a problem [17:38:31] thanks! [17:38:33] ok! [17:39:37] I'll try have a basic implementation up tonight, headed out for a run now [17:39:45] cool! [21:06:05] *** Quits: vin-ivar (~vinit@122.169.11.109) (Read error: No route to host) [21:07:18] *** Joins: vin-ivar (~vinit@122.169.11.109) [21:39:05] MorphDisambiguator.hs: out of memory (requested 2097152 bytes) [21:39:07] hahahaha [21:40:54] so this^ is the largest problem [21:41:11] what i think is [21:41:28] for transferring morph analyses, it's going to be vastly complex because apertium is shallow-transfer [21:41:35] so you have to proceed word by word [21:59:32] also, inariksit, doing a p "el pescado es sabroso" on TranslateSpa.gf gives me like a 2000 and a bit line parse tree [21:59:35] that can't be right, can iit?