BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:A tractable hybrid DDN-POMDP approach to affective dialogue modeli
 ng for probabilistic frame-based dialogue systems - Trung Bui (Twente)
DTSTART:20081020T120000Z
DTEND:20081020T130000Z
UID:TALK14521@talks.cam.ac.uk
CONTACT:Dr Marcus Tomalin
DESCRIPTION:Partially Observable Markov Decision Processes (POMDPs) have b
 een\nempirically demonstrated as a sound model for spoken dialogue managem
 ent\ndesign. This talk will show that the model is also potentially\nappro
 priate for designing affective dialogue problems. \n\nIn the first part of
  my talk\, I will present an affective dialogue model\nusing factored POMD
 Ps and a performance evaluation for this model. Then\,\nI will address sev
 eral practical issues related to the development of\nPOMDP-based dialogue 
 managers for real-world dialogue systems. In\naccordance with researchers 
 in the field\, one of the major problems\nremains tractability.\n\nTo solv
 e this problem\, in the second part\, I will present a hybrid\nDDN-POMDP m
 ethod to developing a tractable affective dialogue model for\nprobabilisti
 c frame-based dialogue systems. The proposed method has two\nnew features:
  (1) being able to deal with a large number of slots & slot\nvalues in rea
 l-time and (2) being able to take into account some aspects\nof the user's
  affective state in deriving the adaptive dialogue\nstrategies. \n\nWe con
 ducted various experiments to evaluate our method and to compare\nit with 
 approximate POMDP techniques and handcrafted policies. The\nresults of the
 se experiments will be discussed in detail.\n
LOCATION:LR5\, Engineering Department\, Baker Building
END:VEVENT
END:VCALENDAR