Contextual Speech Recognition with Difficult Negative Training Examples

Uri Alon; Golan Pundak; Tara N. Sainath

doi:10.1109/ICASSP.2019.8682738

Contextual Speech Recognition with Difficult Negative Training Examples

Uri Alon, Golan Pundak, Tara N. Sainath

Department of Learning and Instructional Sciences

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

Abstract

Improving the representation of contextual information is key to unlocking the potential of end-to-end (E2E) automatic speech recognition (ASR). In this work, we present a novel and simple approach for training an ASR context mechanism with difficult negative examples. The main idea is to focus on proper nouns (e.g., unique entities such as names of people and places) in the reference transcript and use phonetically similar phrases as negative examples, encouraging the neural model to learn more discriminative representations. We apply our approach to an end-to-end contextual ASR model that jointly learns to transcribe and select the correct context items. We show that our proposed method gives up to 53.1% relative improvement in word error rate (WER) across several benchmarks.

Original language	English
Title of host publication	2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	6440-6444
Number of pages	5
ISBN (Electronic)	9781479981311
DOIs	https://doi.org/10.1109/ICASSP.2019.8682738
State	Published - May 2019
Event	44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Brighton, United Kingdom Duration: 12 May 2019 → 17 May 2019

Publication series

Name	ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume	2019-May
ISSN (Print)	1520-6149

Conference

Conference	44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019
Country/Territory	United Kingdom
City	Brighton
Period	12/05/19 → 17/05/19

Bibliographical note

Publisher Copyright:
© 2019 IEEE.

Keywords

attention
biasing
phonetics
sequence-to-sequence models
speech recognition

ASJC Scopus subject areas

Software
Signal Processing
Electrical and Electronic Engineering

Access to Document

10.1109/ICASSP.2019.8682738

Cite this

Alon, U., Pundak, G., & Sainath, T. N. (2019). Contextual Speech Recognition with Difficult Negative Training Examples. In 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings (pp. 6440-6444). Article 8682738 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 2019-May). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICASSP.2019.8682738

Alon, Uri ; Pundak, Golan ; Sainath, Tara N. / Contextual Speech Recognition with Difficult Negative Training Examples. 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2019. pp. 6440-6444 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings).

@inproceedings{3bb83e38c1e74eb9ae49a04066f4d3f7,

title = "Contextual Speech Recognition with Difficult Negative Training Examples",

abstract = "Improving the representation of contextual information is key to unlocking the potential of end-to-end (E2E) automatic speech recognition (ASR). In this work, we present a novel and simple approach for training an ASR context mechanism with difficult negative examples. The main idea is to focus on proper nouns (e.g., unique entities such as names of people and places) in the reference transcript and use phonetically similar phrases as negative examples, encouraging the neural model to learn more discriminative representations. We apply our approach to an end-to-end contextual ASR model that jointly learns to transcribe and select the correct context items. We show that our proposed method gives up to 53.1% relative improvement in word error rate (WER) across several benchmarks.",

keywords = "attention, biasing, phonetics, sequence-to-sequence models, speech recognition",

author = "Uri Alon and Golan Pundak and Sainath, {Tara N.}",

note = "Publisher Copyright: {\textcopyright} 2019 IEEE.; 44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 ; Conference date: 12-05-2019 Through 17-05-2019",

year = "2019",

month = may,

doi = "10.1109/ICASSP.2019.8682738",

language = "English",

series = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "6440--6444",

booktitle = "2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings",

address = "United States",

}

Alon, U, Pundak, G & Sainath, TN 2019, Contextual Speech Recognition with Difficult Negative Training Examples. in 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings., 8682738, ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, vol. 2019-May, Institute of Electrical and Electronics Engineers Inc., pp. 6440-6444, 44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019, Brighton, United Kingdom, 12/05/19. https://doi.org/10.1109/ICASSP.2019.8682738

Contextual Speech Recognition with Difficult Negative Training Examples. / Alon, Uri; Pundak, Golan; Sainath, Tara N.
2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2019. p. 6440-6444 8682738 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 2019-May).

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

TY - GEN

T1 - Contextual Speech Recognition with Difficult Negative Training Examples

AU - Alon, Uri

AU - Pundak, Golan

AU - Sainath, Tara N.

PY - 2019/5

Y1 - 2019/5

N2 - Improving the representation of contextual information is key to unlocking the potential of end-to-end (E2E) automatic speech recognition (ASR). In this work, we present a novel and simple approach for training an ASR context mechanism with difficult negative examples. The main idea is to focus on proper nouns (e.g., unique entities such as names of people and places) in the reference transcript and use phonetically similar phrases as negative examples, encouraging the neural model to learn more discriminative representations. We apply our approach to an end-to-end contextual ASR model that jointly learns to transcribe and select the correct context items. We show that our proposed method gives up to 53.1% relative improvement in word error rate (WER) across several benchmarks.

AB - Improving the representation of contextual information is key to unlocking the potential of end-to-end (E2E) automatic speech recognition (ASR). In this work, we present a novel and simple approach for training an ASR context mechanism with difficult negative examples. The main idea is to focus on proper nouns (e.g., unique entities such as names of people and places) in the reference transcript and use phonetically similar phrases as negative examples, encouraging the neural model to learn more discriminative representations. We apply our approach to an end-to-end contextual ASR model that jointly learns to transcribe and select the correct context items. We show that our proposed method gives up to 53.1% relative improvement in word error rate (WER) across several benchmarks.

KW - attention

KW - biasing

KW - phonetics

KW - sequence-to-sequence models

KW - speech recognition

UR - http://www.scopus.com/inward/record.url?scp=85068971664&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2019.8682738

DO - 10.1109/ICASSP.2019.8682738

M3 - Conference contribution

AN - SCOPUS:85068971664

T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

SP - 6440

EP - 6444

BT - 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019

Y2 - 12 May 2019 through 17 May 2019

ER -

Alon U, Pundak G, Sainath TN. Contextual Speech Recognition with Difficult Negative Training Examples. In 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2019. p. 6440-6444. 8682738. (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings). doi: 10.1109/ICASSP.2019.8682738

Contextual Speech Recognition with Difficult Negative Training Examples

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this