BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Expectation Propagation for POMDP Spoken Dialogue Models - Blaise 
 Thomson\, Dialogue Systems Group\, University of Cambridge
DTSTART:20110512T130000Z
DTEND:20110512T143000Z
UID:TALK31370@talks.cam.ac.uk
CONTACT:David Duvenaud
DESCRIPTION:This talk discusses the application of the partially observabl
 e Markov decision process (POMDP) to spoken dialogue and how the model can
  be used to build a system that interacts with users via speech. The focus
  of the talk will be on the use of the expectation propagation (EP) with v
 arious optimisations to efficiently learn parameters of the POMDP model\, 
 although an overview of the full system will also be provided.  Interestin
 gly\, the EP algorithm provides a way to learn models of how humans behave
  in a dialogue with only a noisy estimate of the semantics of their uttera
 nce. No human annotations of either the state of the dialogue or the seman
 tics are required. An evaluation of the learning algorithm as well as comp
 arisons of the POMDP approach with other techniques will be presented.
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR