BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Neural Belief Tracker: Data-Driven Dialogue State Tracking using S
 emantically Specialised Vector Spaces - Nikola Mrskic\, University of Camb
 ridge
DTSTART:20170512T110000Z
DTEND:20170512T120000Z
UID:TALK72534@talks.cam.ac.uk
CONTACT:Amandla Mabona
DESCRIPTION:One of the core components of modern spoken dialogue systems i
 s the belief tracker\, which estimates the user's goal at every step of th
 e dialogue. However\, most current approaches have difficulty scaling to l
 arger\, more complex dialogue domains. This is due to their dependency on 
 either: a) Spoken Language Understanding models that require large amounts
  of annotated training data\; or b) hand-crafted lexicons for capturing so
 me of the linguistic variation in users' language. We propose a novel Neur
 al Belief Tracking (NBT) framework which overcomes these problems by build
 ing on recent advances in representation learning. NBT models reason over 
 pre-trained\, semantically specialised word vectors\, learning to compose 
 them into distributed representations of user utterances and dialogue cont
 ext. Our evaluation on two datasets shows that this approach surpasses pas
 t limitations\, matching the performance of state-of-the-art models which 
 rely on hand-crafted semantic lexicons and outperforming them when such le
 xicons are not provided. Finally\, we will discuss how the properties of u
 nderlying vector spaces impact model performance\, and how the fact that t
 he proposed model operates purely over word vectors allows immediate deplo
 yment of belief tracking models for other languages.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR