BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Dialog state tracking: Open problems\, challenge task\, and recent
  work - Jason D. Williams\, Microsoft Research
DTSTART:20130611T110000Z
DTEND:20130611T120000Z
UID:TALK45641@talks.cam.ac.uk
CONTACT:Catherine Breslin
DESCRIPTION:In conversational systems\, "dialog state tracking" means infe
 rring the user's goal from the conversation history up to the current turn
 .  System responses are chosen based on the estimated dialog state\, so ac
 curate dialog state tracking is crucial to the performance of conversation
 al systems.  However\, state tracking is non-trivial since input is receiv
 ed via the error-prone processes of speech recognition and language unders
 tanding\, so the system is never sure of what the user has said.  \n\nIn c
 ommercial systems\, dialog state tracking is typically done with simple bu
 t rather effective hand-crafted rules.  Over the past 10 years\, the resea
 rch community has developed several statistical methods for state tracking
 .  These have been shown to improve the performance of conversational syst
 ems in lab settings\, and have recently be tested in public deployments fo
 r the first time.\n\nIn this talk\, I'll explain the dialog state tracking
  problem\, and current solutions in industry and research.  I'll then cove
 r two recent public deployments of statistical state tracking\, which reve
 aled several unanticipated weaknesses in state-of-the-art methods for stat
 e tracking.  This study led to the creation of the "Dialog state tracking 
 challenge"\, in which dialog data and evaluation tools were released\; I'l
 l next cover results from the challenge\, which included 27 state trackers
  from 9 teams.  Finally\, I'll describe some recent work on dialog state t
 racking at Microsoft.
LOCATION:Department of Engineering - LR12
END:VEVENT
END:VCALENDAR