BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Efficient Decoding with Generative Score-Spaces Using the Expectat
 ion Semiring - Rogier van Dalen (University of Cambridge)
DTSTART:20121130T123000Z
DTEND:20121130T133000Z
UID:TALK41166@talks.cam.ac.uk
CONTACT:Catherine Breslin
DESCRIPTION:Speech recognisers are usually based on hidden Markov models (
 HMMs).  These are finite-state models\, with observation vectors condition
 ally independent given the state sequence. This assumption yields efficien
 t algorithms\, but it limits the power of the model. An alternative type o
 f model that allows a wide range of features and dependence structures is 
 a log-linear model. For example\, it can use longer-span\, variable-length
  features. For decoding continuous speech\, however\, the optimal combinat
 ion of segmentation of the utterance into words and word sequence must be 
 found. Features must therefore be extracted for each possible segment of a
 udio. For many types of features\, this becomes slow. In this talk\, I wil
 l discuss how long-span features can be derived from the likelihoods of wo
 rd HMMs. Standard adaptation techniques can then be used. Derivatives of t
 he log-likelihoods\, which break the Markov assumption\, are appended. I w
 ill show how to decode with this specific class of model in quadratic time
 .\n\nFollowed by a mini poster session for papers that will be presented t
 he following week at SLT:\n\n"Policy optimisation of POMDP-based dialogue 
 systems without state space compression"\, Milica Gasic\, Matthew Henderso
 n\, Blaise Thomson\, Pirros Tsiakoulis and Steve Young\n\n"Discriminative 
 Spoken Language Understanding Using Word Confusion Networks"\, Matthew Hen
 derson\, Milica Gasic\, Blaise Thomson\, Pirros Tsiakoulis\, Kai Yu\, and 
 Steve Young\n\n"N-Best Error Simulation for Training Spoken Dialogue Syste
 ms"\, Blaise Thomson\, Milica Gasic\, Matthew Henderson\, Pirros Tsiakouli
 s\, and Steve Young\n\nRefreshments will be provided.
LOCATION:Department of Engineering - LR6
END:VEVENT
END:VCALENDAR