You are here

Natural Speech Technology Annual Meeting 2014

20 May 2014 -  ICOSS, the Interdisciplinary Centre of the Social Sciences, Sheffield 


09:00 - 09:30 - Startup and coffee

9.30 - 11.00 Talks: 

  • Steve Renals - Introduction [slides]
  • Heidi Christensen "Update on the homeService user study" [slides]
  • Gustav Eje Henter "Measuring the perceptual effects of speech synthesis modelling assumptions" [slides]
  • Marcus Tomalin "Automatic and Human Filled Pause Insertion for Speech Synthesis" [slides]
  • Oscar Saz "Asynchronous factorisation with CMLLR transformations" [slides]

11.00 - 12.30 coffee and demos/posters 

  • Heidi Christensen/Iñigo Casanueva - homeService videos (CATCH launch video + videos of homeService users using system)
  • Iñigo Casanueva "Adaptive Speech Recognition and Dialogue Management for Users with Speech Disorders"
  • Charles Fox - Environmental models
  • Penny Karanasou - "Adaptation of Deep Neural Network Acoustic Models Using Factorised I-Vectors" [pdf]
  • Yulan Liu - "Using NN front-ends on far field multiple microphones based speech recognition" [pdf] & "Semi-supervised training of DNN" [pdf]
  • Oscar Saz/ Thomas Hain - webASR for media data
  • Oscar Saz - Acoustic background visualisation
  • Peter Bell - Cross-lingual adaptation with multi-task adaptive networks
  • Liang Lu - "Acoustic Data-driven Pronunciation Lexicon for Large Vocabulary Speech Recognition"
  • Joris Driesen - Sky/BBC subtitling
  • Pierre Lanchantin "Multiple Average Voice based Speech Synthesis"[pdf]
  • Tom Merritt -  "Investigation of limited quality of HMM synthesis”[pdf]
  • Christophe Veaux - "Voice Banking and Voice  Reconstruction"[pdf]
  • Phil Green/ Jonathan Kilgour - English Heritage [pdf]
  • Pawel Swietojanski "Hybrid Acoustic Models for Distant and Multichannel Large Vocabulary Speech Recognition"
  • Xunying Liu - "Paraphrastic Neural Network Language Models" 
  • Oscar Saz - "Using Contextual Information in Joint Factor Eigenspace MLLR for Speech Recognition in Diverse Scenarios" [pdf]
  • Heidi Christensen/Iñigo Casanueva - Automatic Selection of Speakers for Improved Acoustic Modelling: Recognition of Disordered Speech with Sparse Data [pdf]

12.30 - 14.00 Lunch 

14.00 - 15.15 Talks 

  • Charles Fox "Extending LIMABEAM" [slides]
  • Chao Zhang "Standalone Training and Context-Independent Initialisations of Context-Dependent Deep Neural Networks" [slides]
  • Pawel Swietojanski "Channel-wise convolution for multi-microphone distant speech recognition" [slides]
  • Xunying Liu "Efficient training and decoding methods for recurrent neural network language models" [slides]

15.15 - 15.45 afternoon coffee

15.45 - 17.00 Discussions  

  • Clinical
  • Media 
  • New Directions (beyond Deep) 

17.00 - 18:30 advisory board meeting

20:00 dinner at Silversmiths (