Multirate ASR models for phone-class dependent N-best list rescoring

Venkata R. Gadde, Mustafa (Kemal) Sonmez, Horacio Franco

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

Speech comprises a variety of acoustical phenomena occurring at differing rates. Fixed-rate ASR systems assume in effect a constant temporal rate of information flow via incorporating uniform statistics in proportion to a sound's duration. The usual tradeoff window length of 25-30 milliseconds represents a time-frequency resolution compromise, which aims to allow reasonable speed for following changes in the spectral trajectories and sufficient number of samples to estimate the harmonic structure. In this work, we describe a technique to augment a recognizer that uses this compromise with information from multiple-rate spectral models that emphasize either better time or better frequency resolution in order to improve performance. The main idea is to use the hypotheses generated by a fixed-rate recognizer to determine the appropriate model rate for a segment of the speech waveform. This is realized through a technique based on rescoring of N-best lists with acoustical models using different temporal windows by a phone-dependent posterior-like score. We report results on the NIST Evaluation 2002 dataset, and demonstrate that the rescoring method produces word error rate (WER) improvements in a baseline system.

Original languageEnglish (US)
Title of host publicationProceedings of ASRU 2005: 2005 IEEE Automatic Speech Recognition and Understanding Workshop
Pages265-269
Number of pages5
Volume2005
DOIs
StatePublished - 2005
Externally publishedYes
EventASRU 2005: 2005 IEEE Automatic Speech Recognition and Understanding Workshop - Cancun, Mexico
Duration: Nov 27 2005Dec 1 2005

Other

OtherASRU 2005: 2005 IEEE Automatic Speech Recognition and Understanding Workshop
CountryMexico
CityCancun
Period11/27/0512/1/05

Fingerprint

Trajectories
Statistics
Acoustic waves

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Gadde, V. R., Sonmez, M. K., & Franco, H. (2005). Multirate ASR models for phone-class dependent N-best list rescoring. In Proceedings of ASRU 2005: 2005 IEEE Automatic Speech Recognition and Understanding Workshop (Vol. 2005, pp. 265-269). [1566513] https://doi.org/10.1109/ASRU.2005.1566513

Multirate ASR models for phone-class dependent N-best list rescoring. / Gadde, Venkata R.; Sonmez, Mustafa (Kemal); Franco, Horacio.

Proceedings of ASRU 2005: 2005 IEEE Automatic Speech Recognition and Understanding Workshop. Vol. 2005 2005. p. 265-269 1566513.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Gadde, VR, Sonmez, MK & Franco, H 2005, Multirate ASR models for phone-class dependent N-best list rescoring. in Proceedings of ASRU 2005: 2005 IEEE Automatic Speech Recognition and Understanding Workshop. vol. 2005, 1566513, pp. 265-269, ASRU 2005: 2005 IEEE Automatic Speech Recognition and Understanding Workshop, Cancun, Mexico, 11/27/05. https://doi.org/10.1109/ASRU.2005.1566513
Gadde VR, Sonmez MK, Franco H. Multirate ASR models for phone-class dependent N-best list rescoring. In Proceedings of ASRU 2005: 2005 IEEE Automatic Speech Recognition and Understanding Workshop. Vol. 2005. 2005. p. 265-269. 1566513 https://doi.org/10.1109/ASRU.2005.1566513
Gadde, Venkata R. ; Sonmez, Mustafa (Kemal) ; Franco, Horacio. / Multirate ASR models for phone-class dependent N-best list rescoring. Proceedings of ASRU 2005: 2005 IEEE Automatic Speech Recognition and Understanding Workshop. Vol. 2005 2005. pp. 265-269
@inproceedings{9a9bec0dc5ff4d63bb053d1aa46b72ff,
title = "Multirate ASR models for phone-class dependent N-best list rescoring",
abstract = "Speech comprises a variety of acoustical phenomena occurring at differing rates. Fixed-rate ASR systems assume in effect a constant temporal rate of information flow via incorporating uniform statistics in proportion to a sound's duration. The usual tradeoff window length of 25-30 milliseconds represents a time-frequency resolution compromise, which aims to allow reasonable speed for following changes in the spectral trajectories and sufficient number of samples to estimate the harmonic structure. In this work, we describe a technique to augment a recognizer that uses this compromise with information from multiple-rate spectral models that emphasize either better time or better frequency resolution in order to improve performance. The main idea is to use the hypotheses generated by a fixed-rate recognizer to determine the appropriate model rate for a segment of the speech waveform. This is realized through a technique based on rescoring of N-best lists with acoustical models using different temporal windows by a phone-dependent posterior-like score. We report results on the NIST Evaluation 2002 dataset, and demonstrate that the rescoring method produces word error rate (WER) improvements in a baseline system.",
author = "Gadde, {Venkata R.} and Sonmez, {Mustafa (Kemal)} and Horacio Franco",
year = "2005",
doi = "10.1109/ASRU.2005.1566513",
language = "English (US)",
isbn = "0780394798",
volume = "2005",
pages = "265--269",
booktitle = "Proceedings of ASRU 2005: 2005 IEEE Automatic Speech Recognition and Understanding Workshop",

}

TY - GEN

T1 - Multirate ASR models for phone-class dependent N-best list rescoring

AU - Gadde, Venkata R.

AU - Sonmez, Mustafa (Kemal)

AU - Franco, Horacio

PY - 2005

Y1 - 2005

N2 - Speech comprises a variety of acoustical phenomena occurring at differing rates. Fixed-rate ASR systems assume in effect a constant temporal rate of information flow via incorporating uniform statistics in proportion to a sound's duration. The usual tradeoff window length of 25-30 milliseconds represents a time-frequency resolution compromise, which aims to allow reasonable speed for following changes in the spectral trajectories and sufficient number of samples to estimate the harmonic structure. In this work, we describe a technique to augment a recognizer that uses this compromise with information from multiple-rate spectral models that emphasize either better time or better frequency resolution in order to improve performance. The main idea is to use the hypotheses generated by a fixed-rate recognizer to determine the appropriate model rate for a segment of the speech waveform. This is realized through a technique based on rescoring of N-best lists with acoustical models using different temporal windows by a phone-dependent posterior-like score. We report results on the NIST Evaluation 2002 dataset, and demonstrate that the rescoring method produces word error rate (WER) improvements in a baseline system.

AB - Speech comprises a variety of acoustical phenomena occurring at differing rates. Fixed-rate ASR systems assume in effect a constant temporal rate of information flow via incorporating uniform statistics in proportion to a sound's duration. The usual tradeoff window length of 25-30 milliseconds represents a time-frequency resolution compromise, which aims to allow reasonable speed for following changes in the spectral trajectories and sufficient number of samples to estimate the harmonic structure. In this work, we describe a technique to augment a recognizer that uses this compromise with information from multiple-rate spectral models that emphasize either better time or better frequency resolution in order to improve performance. The main idea is to use the hypotheses generated by a fixed-rate recognizer to determine the appropriate model rate for a segment of the speech waveform. This is realized through a technique based on rescoring of N-best lists with acoustical models using different temporal windows by a phone-dependent posterior-like score. We report results on the NIST Evaluation 2002 dataset, and demonstrate that the rescoring method produces word error rate (WER) improvements in a baseline system.

UR - http://www.scopus.com/inward/record.url?scp=33846247945&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33846247945&partnerID=8YFLogxK

U2 - 10.1109/ASRU.2005.1566513

DO - 10.1109/ASRU.2005.1566513

M3 - Conference contribution

AN - SCOPUS:33846247945

SN - 0780394798

SN - 9780780394797

VL - 2005

SP - 265

EP - 269

BT - Proceedings of ASRU 2005: 2005 IEEE Automatic Speech Recognition and Understanding Workshop

ER -