BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Error-Aware Probabilistic Parsing - Jennifer Foster\, Dublin City 
 University
DTSTART:20080509T110000Z
DTEND:20080509T120000Z
UID:TALK10737@talks.cam.ac.uk
CONTACT:Johanna Geiss
DESCRIPTION:Given the human propensity to err\, a parser must be able to p
 roduce\naccurate analyses for sentences which are deviant according to hum
 an\nstandards\, yet which we routinely interpret correctly. State-of-the-\
 nart probabilistic parsers are generally robust to errors\, and they\nwill
  return analyses for most ungrammatical sentences. However\, these\nrobust
  analyses are not necessarily correct because they do not always\nreflect 
 the meanings of the ill-formed sentences. I present a two-\nstage "error-a
 ware" probabilistic parsing architecture which uses two\nversions of a pro
 babilistic parser\, one trained on a normal treebank and\nthe other traine
 d on an automatically created ungrammatical version of the\noriginal treeb
 ank. A binary classifier is used to decide which version to\nemploy. I pre
 sent the results of experiments carried out using the\n"error-aware" proba
 bilistic parsing architecture and the Penn Treebank. I\nalso\npresent the 
 results of experiments carried out with various\ngrammatical/ungrammatical
  classifiers and the BNC.
LOCATION:SW01 Computer Laboratory
END:VEVENT
END:VCALENDAR