Training Augmented Models Using SVMs

Mark J.F. GALES; Martin I. LAYTON

doi:10.1093/ietisy/e89-d.3.892

Training Augmented Models Using SVMs

Mark J.F. GALES, Martin I. LAYTON

Full Text Views

0

Cite this

Summary :

There has been significant interest in developing new forms of acoustic model, in particular models which allow additional dependencies to be represented than those contained within a standard hidden Markov model (HMM). This paper discusses one such class of models, augmented statistical models. Here, a local exponential approximation is made about some point on a base model. This allows additional dependencies within the data to be modelled than are represented in the base distribution. Augmented models based on Gaussian mixture models (GMMs) and HMMs are briefly described. These augmented models are then related to generative kernels, one approach used for allowing support vector machines (SVMs) to be applied to variable length data. The training of augmented statistical models within an SVM, generative kernel, framework is then discussed. This may be viewed as using maximum margin training to estimate statistical models. Augmented Gaussian mixture models are then evaluated using rescoring on a large vocabulary speech recognition task.

Publication: IEICE TRANSACTIONS on Information Vol.E89-D No.3 pp.892-899

Publication Date: 2006/03/01

Publicized

Online ISSN: 1745-1361

DOI: 10.1093/ietisy/e89-d.3.892

Type of Manuscript: Special Section INVITED PAPER (Special Section on Statistical Modeling for Speech Processing)

Category

Cite this

Copy

Mark J.F. GALES, Martin I. LAYTON, "Training Augmented Models Using SVMs" in IEICE TRANSACTIONS on Information, vol. E89-D, no. 3, pp. 892-899, March 2006, doi: 10.1093/ietisy/e89-d.3.892.
Abstract: There has been significant interest in developing new forms of acoustic model, in particular models which allow additional dependencies to be represented than those contained within a standard hidden Markov model (HMM). This paper discusses one such class of models, augmented statistical models. Here, a local exponential approximation is made about some point on a base model. This allows additional dependencies within the data to be modelled than are represented in the base distribution. Augmented models based on Gaussian mixture models (GMMs) and HMMs are briefly described. These augmented models are then related to generative kernels, one approach used for allowing support vector machines (SVMs) to be applied to variable length data. The training of augmented statistical models within an SVM, generative kernel, framework is then discussed. This may be viewed as using maximum margin training to estimate statistical models. Augmented Gaussian mixture models are then evaluated using rescoring on a large vocabulary speech recognition task.
URL: https://global.ieice.org/en_transactions/information/10.1093/ietisy/e89-d.3.892/_p

Copy

@ARTICLE{e89-d_3_892,
author={Mark J.F. GALES, Martin I. LAYTON, },
journal={IEICE TRANSACTIONS on Information},
title={Training Augmented Models Using SVMs},
year={2006},
volume={E89-D},
number={3},
pages={892-899},
abstract={There has been significant interest in developing new forms of acoustic model, in particular models which allow additional dependencies to be represented than those contained within a standard hidden Markov model (HMM). This paper discusses one such class of models, augmented statistical models. Here, a local exponential approximation is made about some point on a base model. This allows additional dependencies within the data to be modelled than are represented in the base distribution. Augmented models based on Gaussian mixture models (GMMs) and HMMs are briefly described. These augmented models are then related to generative kernels, one approach used for allowing support vector machines (SVMs) to be applied to variable length data. The training of augmented statistical models within an SVM, generative kernel, framework is then discussed. This may be viewed as using maximum margin training to estimate statistical models. Augmented Gaussian mixture models are then evaluated using rescoring on a large vocabulary speech recognition task.},
keywords={},
doi={10.1093/ietisy/e89-d.3.892},
ISSN={1745-1361},
month={March},}

Copy

TY - JOUR
TI - Training Augmented Models Using SVMs
T2 - IEICE TRANSACTIONS on Information
SP - 892
EP - 899
AU - Mark J.F. GALES
AU - Martin I. LAYTON
PY - 2006
DO - 10.1093/ietisy/e89-d.3.892
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E89-D
IS - 3
JA - IEICE TRANSACTIONS on Information
Y1 - March 2006
AB - There has been significant interest in developing new forms of acoustic model, in particular models which allow additional dependencies to be represented than those contained within a standard hidden Markov model (HMM). This paper discusses one such class of models, augmented statistical models. Here, a local exponential approximation is made about some point on a base model. This allows additional dependencies within the data to be modelled than are represented in the base distribution. Augmented models based on Gaussian mixture models (GMMs) and HMMs are briefly described. These augmented models are then related to generative kernels, one approach used for allowing support vector machines (SVMs) to be applied to variable length data. The training of augmented statistical models within an SVM, generative kernel, framework is then discussed. This may be viewed as using maximum margin training to estimate statistical models. Augmented Gaussian mixture models are then evaluated using rescoring on a large vocabulary speech recognition task.
ER -