BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Improved Parsing and POS Tagging Using Inter-Sentence Consistency 
 Constraints - Roi Reichart\, University of Cambridge
DTSTART:20121123T120000Z
DTEND:20121123T130000Z
UID:TALK41259@talks.cam.ac.uk
CONTACT:Ekaterina Kochmar
DESCRIPTION:State-of-the-art statistical parsers and POS taggers perform v
 ery well\nwhen trained with large amounts of in-domain data. When training
  data\nis out-of-domain or limited\, accuracy degrades. In this work\, we 
 aim\nto compensate for the lack of available training data by exploiting\n
 similarities between test set sentences. We show how to augment\nsentence 
 level models for parsing and POS tagging with inter-sentence\nconsistency 
 constraints. To deal with the resulting global objective\,\nwe present an 
 efficient and exact dual decomposition decoding\nalgorithm. In experiments
 \, we add\nconsistency constraints to the MST parser and the Stanford\npar
 t-of-speech tagger and demonstrate significant error reduction in\nthe dom
 ain adaptation and the lightly supervised settings across five\nlanguages.
 \n\nJoint work with Alexander Rush\, Michael Collins and Amir Globerson
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR