Smoothing Method for Improved Minimum Phone Error Linear Regression
Summary :
A smoothing method for minimum phone error linear regression (MPELR) is proposed in this paper. We show that the objective function for minimum phone error (MPE) can be combined with a prior mean distribution. When the prior mean distribution is based on maximum likelihood (ML) estimates, the proposed method is the same as the previous smoothing technique for MPELR. Instead of ML estimates, maximum a posteriori (MAP) parameter estimate is used to define the mode of prior mean distribution to improve the performance of MPELR. Experiments on a large vocabulary speech recognition task show that the proposed method can obtain 8.4% relative reduction in word error rate when the amount of data is limited, while retaining the same asymptotic performance as conventional MPELR. When compared with discriminative maximum a posteriori linear regression (DMAPLR), the proposed method shows improvement except for the case of limited adaptation data for supervised adaptation.
- Publication
- IEICE TRANSACTIONS on Information Vol.E97-D No.8 pp.2105-2113
- Publication Date
- 2014/08/01
- Publicized
- Online ISSN
- 1745-1361
- DOI
- 10.1587/transinf.E97.D.2105
- Type of Manuscript
- PAPER
- Category
- Speech and Hearing
Authors
Yaohui QI
Beijing Institute of Technology,Chinese Academy of Sciences,Hebei Normal University
Fuping PAN
Chinese Academy of Sciences
Fengpei GE
Chinese Academy of Sciences
Qingwei ZHAO
Chinese Academy of Sciences
Yonghong YAN
Beijing Institute of Technology,Chinese Academy of Sciences
Keyword
Latest Issue
Copyrights notice of machine-translated contents
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Cite this
Copy
Yaohui QI, Fuping PAN, Fengpei GE, Qingwei ZHAO, Yonghong YAN, "Smoothing Method for Improved Minimum Phone Error Linear Regression" in IEICE TRANSACTIONS on Information,
vol. E97-D, no. 8, pp. 2105-2113, August 2014, doi: 10.1587/transinf.E97.D.2105.
Abstract: A smoothing method for minimum phone error linear regression (MPELR) is proposed in this paper. We show that the objective function for minimum phone error (MPE) can be combined with a prior mean distribution. When the prior mean distribution is based on maximum likelihood (ML) estimates, the proposed method is the same as the previous smoothing technique for MPELR. Instead of ML estimates, maximum a posteriori (MAP) parameter estimate is used to define the mode of prior mean distribution to improve the performance of MPELR. Experiments on a large vocabulary speech recognition task show that the proposed method can obtain 8.4% relative reduction in word error rate when the amount of data is limited, while retaining the same asymptotic performance as conventional MPELR. When compared with discriminative maximum a posteriori linear regression (DMAPLR), the proposed method shows improvement except for the case of limited adaptation data for supervised adaptation.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.E97.D.2105/_p
Copy
@ARTICLE{e97-d_8_2105,
author={Yaohui QI, Fuping PAN, Fengpei GE, Qingwei ZHAO, Yonghong YAN, },
journal={IEICE TRANSACTIONS on Information},
title={Smoothing Method for Improved Minimum Phone Error Linear Regression},
year={2014},
volume={E97-D},
number={8},
pages={2105-2113},
abstract={A smoothing method for minimum phone error linear regression (MPELR) is proposed in this paper. We show that the objective function for minimum phone error (MPE) can be combined with a prior mean distribution. When the prior mean distribution is based on maximum likelihood (ML) estimates, the proposed method is the same as the previous smoothing technique for MPELR. Instead of ML estimates, maximum a posteriori (MAP) parameter estimate is used to define the mode of prior mean distribution to improve the performance of MPELR. Experiments on a large vocabulary speech recognition task show that the proposed method can obtain 8.4% relative reduction in word error rate when the amount of data is limited, while retaining the same asymptotic performance as conventional MPELR. When compared with discriminative maximum a posteriori linear regression (DMAPLR), the proposed method shows improvement except for the case of limited adaptation data for supervised adaptation.},
keywords={},
doi={10.1587/transinf.E97.D.2105},
ISSN={1745-1361},
month={August},}
Copy
TY - JOUR
TI - Smoothing Method for Improved Minimum Phone Error Linear Regression
T2 - IEICE TRANSACTIONS on Information
SP - 2105
EP - 2113
AU - Yaohui QI
AU - Fuping PAN
AU - Fengpei GE
AU - Qingwei ZHAO
AU - Yonghong YAN
PY - 2014
DO - 10.1587/transinf.E97.D.2105
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E97-D
IS - 8
JA - IEICE TRANSACTIONS on Information
Y1 - August 2014
AB - A smoothing method for minimum phone error linear regression (MPELR) is proposed in this paper. We show that the objective function for minimum phone error (MPE) can be combined with a prior mean distribution. When the prior mean distribution is based on maximum likelihood (ML) estimates, the proposed method is the same as the previous smoothing technique for MPELR. Instead of ML estimates, maximum a posteriori (MAP) parameter estimate is used to define the mode of prior mean distribution to improve the performance of MPELR. Experiments on a large vocabulary speech recognition task show that the proposed method can obtain 8.4% relative reduction in word error rate when the amount of data is limited, while retaining the same asymptotic performance as conventional MPELR. When compared with discriminative maximum a posteriori linear regression (DMAPLR), the proposed method shows improvement except for the case of limited adaptation data for supervised adaptation.
ER -
English
