BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Shift-Reduce CCG Parsing with a Dependency Model - Wenduan Xu\, Un
 iversity of Cambridge
DTSTART:20140605T120000Z
DTEND:20140605T123000Z
UID:TALK52870@talks.cam.ac.uk
CONTACT:Ekaterina Kochmar
DESCRIPTION:We present the first dependency model for a shift-reduce CCG p
 arser.\nModelling dependencies is desirable for a number of reasons\, incl
 uding\nhandling the “spurious” ambiguity of CCG\; fitting well with th
 e theory\nof CCG\; and optimizing for structures which are evaluated at te
 st\ntime. We develop a novel training technique using a dependency oracle\
 ,\nin which all derivations are hidden. A challenge arises from the fact\n
 that the oracle needs to keep track of exponentially many\ngold-standard d
 erivations\, which is solved by integrating a packed\nparse forest with th
 e beam-search decoder. Standard CCGBank tests show\nthe model achieves up 
 to 1.05 labeled F-score improvements over three\nexisting\, competitive CC
 G parsing models.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR