Weakly supervised spoken term discovery using cross-lingual side information

Sameer Bansal; Herman Kamper; Sharon Goldwater; Adam Lopez

doi:10.1109/ICASSP.2017.7953260

Weakly supervised spoken term discovery using cross-lingual side information

Sameer Bansal, Herman Kamper, Sharon Goldwater, Adam Lopez

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

Abstract

Recent work on unsupervised term discovery (UTD) aims to identify and cluster repeated word-like units from audio alone. These systems are promising for some very low-resource languages where transcribed audio is unavailable, or where no written form of the language exists. However, in some cases it may still be feasible (e.g., through crowdsourcing) to obtain (possibly noisy) text translations of the audio. If so, this information could be used as a source of side information to improve UTD. Here, we present a simple method for rescoring the output of a UTD system using text translations, and test it on a corpus of Spanish audio with English translations. We show that it greatly improves the average precision of the results over a wide range of system configurations and data preprocessing methods.

Original language	English
Title of host publication	The 42nd IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2017)
Publisher	Institute of Electrical and Electronics Engineers (IEEE)
Pages	5760-5764
Number of pages	5
ISBN (Electronic)	978-1-5090-4117-6
DOIs	https://doi.org/10.1109/ICASSP.2017.7953260
Publication status	Published - 19 Jun 2017
Event	42nd IEEE International Conference on Acoustics, Speech and Signal Processing - New Orleans, United States Duration: 5 Mar 2017 → 9 Mar 2017 http://www.ieee-icassp2017.org/

Publication series

Name
Publisher	IEEE
ISSN (Electronic)	2379-190X

Conference

Conference	42nd IEEE International Conference on Acoustics, Speech and Signal Processing
Abbreviated title	ICASSP 2017
Country/Territory	United States
City	New Orleans
Period	5/03/17 → 9/03/17
Internet address	http://www.ieee-icassp2017.org/

Access to Document

10.1109/ICASSP.2017.7953260

1609.06530v1Accepted author manuscript, 910 KB

http://ieeexplore.ieee.org/document/7953260/

Cite this

@inproceedings{001804d40fe94bc0b02e72f421742ce5,

title = "Weakly supervised spoken term discovery using cross-lingual side information",

abstract = "Recent work on unsupervised term discovery (UTD) aims to identify and cluster repeated word-like units from audio alone. These systems are promising for some very low-resource languages where transcribed audio is unavailable, or where no written form of the language exists. However, in some cases it may still be feasible (e.g., through crowdsourcing) to obtain (possibly noisy) text translations of the audio. If so, this information could be used as a source of side information to improve UTD. Here, we present a simple method for rescoring the output of a UTD system using text translations, and test it on a corpus of Spanish audio with English translations. We show that it greatly improves the average precision of the results over a wide range of system configurations and data preprocessing methods.",

author = "Sameer Bansal and Herman Kamper and Sharon Goldwater and Adam Lopez",

year = "2017",

month = jun,

day = "19",

doi = "10.1109/ICASSP.2017.7953260",

language = "English",

publisher = "Institute of Electrical and Electronics Engineers (IEEE)",

pages = "5760--5764",

booktitle = "The 42nd IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2017)",

address = "United States",

note = "42nd IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2017 ; Conference date: 05-03-2017 Through 09-03-2017",

url = "http://www.ieee-icassp2017.org/",

}

Bansal, S, Kamper, H, Goldwater, S & Lopez, A 2017, Weakly supervised spoken term discovery using cross-lingual side information. in The 42nd IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2017). Institute of Electrical and Electronics Engineers (IEEE), pp. 5760-5764, 42nd IEEE International Conference on Acoustics, Speech and Signal Processing, New Orleans, Louisiana, United States, 5/03/17. https://doi.org/10.1109/ICASSP.2017.7953260

Weakly supervised spoken term discovery using cross-lingual side information. / Bansal, Sameer; Kamper, Herman; Goldwater, Sharon et al.

The 42nd IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2017). Institute of Electrical and Electronics Engineers (IEEE), 2017. p. 5760-5764.

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

TY - GEN

T1 - Weakly supervised spoken term discovery using cross-lingual side information

AU - Bansal, Sameer

AU - Kamper, Herman

AU - Goldwater, Sharon

AU - Lopez, Adam

PY - 2017/6/19

Y1 - 2017/6/19

N2 - Recent work on unsupervised term discovery (UTD) aims to identify and cluster repeated word-like units from audio alone. These systems are promising for some very low-resource languages where transcribed audio is unavailable, or where no written form of the language exists. However, in some cases it may still be feasible (e.g., through crowdsourcing) to obtain (possibly noisy) text translations of the audio. If so, this information could be used as a source of side information to improve UTD. Here, we present a simple method for rescoring the output of a UTD system using text translations, and test it on a corpus of Spanish audio with English translations. We show that it greatly improves the average precision of the results over a wide range of system configurations and data preprocessing methods.

AB - Recent work on unsupervised term discovery (UTD) aims to identify and cluster repeated word-like units from audio alone. These systems are promising for some very low-resource languages where transcribed audio is unavailable, or where no written form of the language exists. However, in some cases it may still be feasible (e.g., through crowdsourcing) to obtain (possibly noisy) text translations of the audio. If so, this information could be used as a source of side information to improve UTD. Here, we present a simple method for rescoring the output of a UTD system using text translations, and test it on a corpus of Spanish audio with English translations. We show that it greatly improves the average precision of the results over a wide range of system configurations and data preprocessing methods.

U2 - 10.1109/ICASSP.2017.7953260

DO - 10.1109/ICASSP.2017.7953260

M3 - Conference contribution

SP - 5760

EP - 5764

BT - The 42nd IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2017)

PB - Institute of Electrical and Electronics Engineers (IEEE)

T2 - 42nd IEEE International Conference on Acoustics, Speech and Signal Processing

Y2 - 5 March 2017 through 9 March 2017

ER -

Weakly supervised spoken term discovery using cross-lingual side information

Abstract

Publication series

Conference

Access to Document

Fingerprint

Cite this