The UEDIN ASR Systems for the IWSLT 2014 Evaluation

Peter Bell; Pawel Swietojanski; Joris Driesen; Mark Sinclair; Fergus McInnes; Steve Renals

The UEDIN ASR Systems for the IWSLT 2014 Evaluation

Peter Bell, Pawel Swietojanski, Joris Driesen, Mark Sinclair, Fergus McInnes, Steve Renals

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract / Description of output

This paper describes the University of Edinburgh (UEDIN) ASR systems for the 2014 IWSLT Evaluation. Notable features of the English system include deep neural network acoustic models in both tandem and hybrid configuration with the use of multi-level adaptive networks, LHUC adaptation and Maxout units. The German system includes lightly supervised training and a new method for dictionary generation. Our voice activity detection system now uses a semi-Markov model to incorporate a prior on utterance lengths. There are improvements of up to 30% relative WER on the tst2013 English test set.

Original language	English
Title of host publication	11th International Workshop on Spoken Language Translation (IWSLT 2014)
Number of pages	8
Publication status	Published - 2014

Access to Document

belliwst2014Accepted author manuscript, 180 KB

EU-Bridge
Renals, S., King, S., Koehn, P. & Osborne, M.
EU government bodies
1/02/12 → 31/01/15
Project: Research
inEVENT
Renals, S.
EU government bodies
1/11/11 → 31/10/14
Project: Research
Natural Speech Technology
Renals, S. & King, S.
EPSRC
1/05/11 → 31/07/16
Project: Research

Cite this

@inproceedings{903f2107febc47c3bb68f6ab43f48af6,

title = "The UEDIN ASR Systems for the IWSLT 2014 Evaluation",

abstract = "This paper describes the University of Edinburgh (UEDIN) ASR systems for the 2014 IWSLT Evaluation. Notable features of the English system include deep neural network acoustic models in both tandem and hybrid configuration with the use of multi-level adaptive networks, LHUC adaptation and Maxout units. The German system includes lightly supervised training and a new method for dictionary generation. Our voice activity detection system now uses a semi-Markov model to incorporate a prior on utterance lengths. There are improvements of up to 30% relative WER on the tst2013 English test set.",

author = "Peter Bell and Pawel Swietojanski and Joris Driesen and Mark Sinclair and Fergus McInnes and Steve Renals",

year = "2014",

language = "English",

booktitle = "11th International Workshop on Spoken Language Translation (IWSLT 2014)",

}

TY - GEN

T1 - The UEDIN ASR Systems for the IWSLT 2014 Evaluation

AU - Bell, Peter

AU - Swietojanski, Pawel

AU - Driesen, Joris

AU - Sinclair, Mark

AU - McInnes, Fergus

AU - Renals, Steve

PY - 2014

Y1 - 2014

N2 - This paper describes the University of Edinburgh (UEDIN) ASR systems for the 2014 IWSLT Evaluation. Notable features of the English system include deep neural network acoustic models in both tandem and hybrid configuration with the use of multi-level adaptive networks, LHUC adaptation and Maxout units. The German system includes lightly supervised training and a new method for dictionary generation. Our voice activity detection system now uses a semi-Markov model to incorporate a prior on utterance lengths. There are improvements of up to 30% relative WER on the tst2013 English test set.

AB - This paper describes the University of Edinburgh (UEDIN) ASR systems for the 2014 IWSLT Evaluation. Notable features of the English system include deep neural network acoustic models in both tandem and hybrid configuration with the use of multi-level adaptive networks, LHUC adaptation and Maxout units. The German system includes lightly supervised training and a new method for dictionary generation. Our voice activity detection system now uses a semi-Markov model to incorporate a prior on utterance lengths. There are improvements of up to 30% relative WER on the tst2013 English test set.

M3 - Conference contribution

BT - 11th International Workshop on Spoken Language Translation (IWSLT 2014)

ER -

The UEDIN ASR Systems for the IWSLT 2014 Evaluation

Abstract / Description of output

Access to Document

Fingerprint

Projects

EU-Bridge

inEVENT

Natural Speech Technology

Cite this