[00:17:07] inariksit, won't be coming this year D: i need to go get a bunch of stuff from home for my visa [00:17:09] but i will try to make it to SA [00:41:01] *** Quits: stoopkid (uid137696@gateway/web/irccloud.com/x-wqjwkitpminoqray) (Quit: Connection closed for inactivity) [03:32:20] *** Quits: proteusguy (~proteus-g@2405:9800:b408:bc31:ac62:e116:cecf:20cf) (Remote host closed the connection) [04:22:08] *** Joins: stoopkid (uid137696@gateway/web/irccloud.com/x-iqkygugxwbkmxkjw) [10:13:25] *** Joins: sirdancealot (~sirdancea@79.98.72.195) [11:32:13] *** Joins: fotonzade (~fotonzade@88.244.244.93) [12:06:00] *** Quits: fotonzade (~fotonzade@88.244.244.93) (Ping timeout: 260 seconds) [14:30:52] *** Joins: fotonzade (~fotonzade@88.244.244.93) [15:59:53] *** Joins: keri (~pepijn@ksit.fixme.fi) [16:03:54] hi! can anyone explain what PGF the wide coverage demo is running, and how I can build it myself? [16:04:23] (and also what else the wide coverage demo is doing, e.g. w.r.t. tokenization and such) [16:05:52] because the regular Translate8 grammar seems incredibly brittle w.r.t. punctuation and words, whereas the wide coverage demo seems... well, a lot more wide coverage ^^ [16:06:55] (i'm talking about this http://cloud.grammaticalframework.org/wc.html versus the code in GF/lib/src/translator) [17:29:11] *** Joins: proteusguy (~proteus-g@2405:9800:b408:bc31:71df:9c43:18f9:297b) [17:46:25] *** Quits: fotonzade (~fotonzade@88.244.244.93) (Ping timeout: 240 seconds) [19:24:40] keri: yeah that is a mystery for me too >__> [19:26:22] http://school.grammaticalframework.org/2015/#schedule for Monday 20 July the title "The mysteries of the wide-coverage translation system [Thomas]", I remember seeing some slides about that later [19:26:31] but there's no link [19:28:15] * inariksit searches for tokenisation in the gf-dev group [19:28:48] nothing newer than 2014 [19:31:53] are you using the haskell runtime, or the C runtime via haskell bindings? [20:03:38] I bumped the thread from 2014, let's see if anyone answers :-P this is krasimir's answer from 2014: [20:04:18] "In the new C based GF runtime the tokenization is done by the parser itself and then the grammar can control whether and where to put space. Also the &+ hack is not needed anymore. This is not implemented in the pure Haskell runtime. " [20:46:37] *** Joins: fotonzade (~fotonzade@78.189.158.97) [21:02:01] *** Quits: vin-ivar (~vin-ivar@unaffiliated/vin-ivar) (Quit: Later) [21:06:55] *** Quits: fotonzade (~fotonzade@78.189.158.97) (Quit: Leaving) [21:41:59] *** Joins: vin-ivar (~vin-ivar@unaffiliated/vin-ivar) [23:10:25] *** Quits: sirdancealot (~sirdancea@79.98.72.195) (Ping timeout: 240 seconds) [23:15:25] *** Quits: proteusguy (~proteus-g@2405:9800:b408:bc31:71df:9c43:18f9:297b) (Remote host closed the connection)