Active Learning Using Phone-Error Distribution for Speech Modeling

Hiroko MURAKAMI; Koichi SHINODA; Sadaoki FURUI

doi:10.1587/transinf.E95.D.2486

Active Learning Using Phone-Error Distribution for Speech Modeling

Hiroko MURAKAMI, Koichi SHINODA, Sadaoki FURUI

Full Text Views

0

Share
Cite this

Summary :

We propose an active learning framework for speech recognition that reduces the amount of data required for acoustic modeling. This framework consists of two steps. We first obtain a phone-error distribution using an acoustic model estimated from transcribed speech data. Then, from a text corpus we select a sentence whose phone-occurrence distribution is close to the phone-error distribution and collect its speech data. We repeat this process to increase the amount of transcribed speech data. We applied this framework to speaker adaptation and acoustic model training. Our evaluation results showed that it significantly reduced the amount of transcribed data while maintaining the same level of accuracy.

Publication: IEICE TRANSACTIONS on Information Vol.E95-D No.10 pp.2486-2494

Publication Date: 2012/10/01

Publicized

Online ISSN: 1745-1361

DOI: 10.1587/transinf.E95.D.2486

Type of Manuscript: PAPER

Category: Speech and Hearing

Cite this

Copy

Hiroko MURAKAMI, Koichi SHINODA, Sadaoki FURUI, "Active Learning Using Phone-Error Distribution for Speech Modeling" in IEICE TRANSACTIONS on Information, vol. E95-D, no. 10, pp. 2486-2494, October 2012, doi: 10.1587/transinf.E95.D.2486.
Abstract: We propose an active learning framework for speech recognition that reduces the amount of data required for acoustic modeling. This framework consists of two steps. We first obtain a phone-error distribution using an acoustic model estimated from transcribed speech data. Then, from a text corpus we select a sentence whose phone-occurrence distribution is close to the phone-error distribution and collect its speech data. We repeat this process to increase the amount of transcribed speech data. We applied this framework to speaker adaptation and acoustic model training. Our evaluation results showed that it significantly reduced the amount of transcribed data while maintaining the same level of accuracy.
URL: https://globals.ieice.org/en_transactions/information/10.1587/transinf.E95.D.2486/_p

Copy

@ARTICLE{e95-d_10_2486,
author={Hiroko MURAKAMI, Koichi SHINODA, Sadaoki FURUI, },
journal={IEICE TRANSACTIONS on Information},
title={Active Learning Using Phone-Error Distribution for Speech Modeling},
year={2012},
volume={E95-D},
number={10},
pages={2486-2494},
abstract={We propose an active learning framework for speech recognition that reduces the amount of data required for acoustic modeling. This framework consists of two steps. We first obtain a phone-error distribution using an acoustic model estimated from transcribed speech data. Then, from a text corpus we select a sentence whose phone-occurrence distribution is close to the phone-error distribution and collect its speech data. We repeat this process to increase the amount of transcribed speech data. We applied this framework to speaker adaptation and acoustic model training. Our evaluation results showed that it significantly reduced the amount of transcribed data while maintaining the same level of accuracy.},
keywords={},
doi={10.1587/transinf.E95.D.2486},
ISSN={1745-1361},
month={October},}

Copy

TY - JOUR
TI - Active Learning Using Phone-Error Distribution for Speech Modeling
T2 - IEICE TRANSACTIONS on Information
SP - 2486
EP - 2494
AU - Hiroko MURAKAMI
AU - Koichi SHINODA
AU - Sadaoki FURUI
PY - 2012
DO - 10.1587/transinf.E95.D.2486
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E95-D
IS - 10
JA - IEICE TRANSACTIONS on Information
Y1 - October 2012
AB - We propose an active learning framework for speech recognition that reduces the amount of data required for acoustic modeling. This framework consists of two steps. We first obtain a phone-error distribution using an acoustic model estimated from transcribed speech data. Then, from a text corpus we select a sentence whose phone-occurrence distribution is close to the phone-error distribution and collect its speech data. We repeat this process to increase the amount of transcribed speech data. We applied this framework to speaker adaptation and acoustic model training. Our evaluation results showed that it significantly reduced the amount of transcribed data while maintaining the same level of accuracy.
ER -