BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:The Geometry of Machine Translation - Rory Waite (University of Ca
 mbridge)
DTSTART:20150116T133000Z
DTEND:20150116T143000Z
UID:TALK56782@talks.cam.ac.uk
CONTACT:Rogier van Dalen
DESCRIPTION:Most modern statistical machine translation systems are based 
 on linear statistical models.  One extremely effective method for estimati
 ng the model parameters is minimum error rate training (MERT)\, which is a
 n efficient form of line search adapted to the highly non-linear objective
  functions used in machine translation.   We will show that MERT can be re
 presented using convex geometry\, which is the mathematics of polytopes an
 d their faces.   Using this geometric representation of MERT we investigat
 e whether the optimisation of linear models is tractable in general.  It h
 as been believed that the number of feasible solutions of a linear model i
 s exponential with respect to the number of sentences used for parameter e
 stimation\, however we show that the exponential complexity is instead due
  to the feature dimension.   This result has important ramifications becau
 se it suggests that the current trend in building statistical machine tran
 slation systems by introducing very large number of sparse features is inh
 erently not robust.\n\n*Biography*\n\nRory is a research assistant and a r
 ecent graduate student from the University of Cambridge. His research is i
 n statistical machine translation.
LOCATION:Department of Engineering - LR6
END:VEVENT
END:VCALENDAR