BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Helping computers talk from experience - Blaise Thomson\, Universi
 ty of Cambridge
DTSTART:20120323T110000Z
DTEND:20120323T120000Z
UID:TALK36170@talks.cam.ac.uk
CONTACT:Ekaterina Kochmar
DESCRIPTION:In many applications\, spoken dialogue is a compelling method 
 for\ninteracting with computers. With the popularity of mobile devices\, v
 oice\ninterfaces are becoming increasingly important\, but the technology 
 for\nbuilding these interfaces is often very poor. This talk will discuss 
 how\nstatistical methods can aid in the decision making processes of these
 \nspoken dialogue systems. In particular\, we will discuss how Expectation
 \nPropagation (EP) can be used to build models of user behaviour in spoken
 \ndialogues and how reinforcement learning can be used to optimise the\nde
 cision making. EP provides an efficient way to train the parameters and\nu
 pdate the beliefs of a spoken dialogue systems based on the partially\nobs
 ervable Markov decision process. These parameters can even be learned\nusi
 ng noisy observations\, and do not require any annotations besides\nsemant
 ic representations of the speech recognition output of a dialogue.\nThe re
 sulting systems are shown to be more robust to errors than standard\nappro
 aches\, largely because the models are able to handle the uncertainty\nin 
 the dialogue in a principled way.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR