Cross-Lingual Topic Prediction for Speech Using Translations

Sameer Bansal; Herman Kamper; Adam Lopez; Sharon Goldwater

doi:10.1109/ICASSP40776.2020.9054169

Cross-Lingual Topic Prediction for Speech Using Translations

Sameer Bansal, Herman Kamper, Adam Lopez, Sharon Goldwater

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

Abstract

Given a large amount of unannotated speech in a low-resource language, can we classify the speech utterances by topic? We consider this question in the setting where a small amount of speech in the low-resource language is paired with text translations in a high-resource language. We develop an effective cross-lingual topic classifier by training on just 20 hours of translated speech, using a recent model for direct speech-to-text translation. While the translations are poor, they are still good enough to correctly classify the topic of 1-minute speech segments over 70% of the time—a 20% improvement over a majority-class baseline. Such a system could be useful for humanitarian applications like crisis response, where incoming speech in a foreign low-resource language must be quickly assessed for further action.

Original language	English
Title of host publication	ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
Publisher	Institute of Electrical and Electronics Engineers (IEEE)
Pages	8164-8168
Number of pages	5
ISBN (Electronic)	978-1-5090-6631-5
ISBN (Print)	978-1-5090-6632-2
DOIs	https://doi.org/10.1109/ICASSP40776.2020.9054169
Publication status	Published - 14 May 2020
Event	2020 IEEE International Conference on Acoustics, Speech, and Signal Processing - Barcelona, Spain Duration: 4 May 2020 → 8 May 2020 Conference number: 45

Publication series

Name
Publisher	IEEE
ISSN (Print)	1520-6149
ISSN (Electronic)	2379-190X

Conference

Conference	2020 IEEE International Conference on Acoustics, Speech, and Signal Processing
Abbreviated title	ICASSP 2020
Country/Territory	Spain
City	Barcelona
Period	4/05/20 → 8/05/20

Keywords

speech translation
Low-resource speech processing
speech classification
unwritten languages

Access to Document

10.1109/ICASSP40776.2020.9054169Licence: All Rights Reserved

Cross-Lingual Topic Prediction_BANSAL_DOA24012020_AFVAccepted author manuscript, 329 KB

https://ieeexplore.ieee.org/document/9054169Licence: All Rights Reserved

Cite this

@inproceedings{e4cd3f75305f41aa9bef2d5dbc860cc8,

title = "Cross-Lingual Topic Prediction for Speech Using Translations",

abstract = "Given a large amount of unannotated speech in a low-resource language, can we classify the speech utterances by topic? We consider this question in the setting where a small amount of speech in the low-resource language is paired with text translations in a high-resource language. We develop an effective cross-lingual topic classifier by training on just 20 hours of translated speech, using a recent model for direct speech-to-text translation. While the translations are poor, they are still good enough to correctly classify the topic of 1-minute speech segments over 70% of the time—a 20% improvement over a majority-class baseline. Such a system could be useful for humanitarian applications like crisis response, where incoming speech in a foreign low-resource language must be quickly assessed for further action.",

keywords = "speech translation, Low-resource speech processing, speech classification, unwritten languages",

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

year = "2020",

month = may,

day = "14",

doi = "10.1109/ICASSP40776.2020.9054169",

language = "English",

isbn = "978-1-5090-6632-2",

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

pages = "8164--8168",

booktitle = "ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)",

address = "United States",

note = "2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2020 ; Conference date: 04-05-2020 Through 08-05-2020",

}

Bansal, S, Kamper, H, Lopez, A & Goldwater, S 2020, Cross-Lingual Topic Prediction for Speech Using Translations. in ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Institute of Electrical and Electronics Engineers (IEEE), pp. 8164-8168, 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing, Barcelona, Spain, 4/05/20. https://doi.org/10.1109/ICASSP40776.2020.9054169

Cross-Lingual Topic Prediction for Speech Using Translations. / Bansal, Sameer; Kamper, Herman; Lopez, Adam et al.

ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Institute of Electrical and Electronics Engineers (IEEE), 2020. p. 8164-8168.

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

TY - GEN

T1 - Cross-Lingual Topic Prediction for Speech Using Translations

AU - Bansal, Sameer

AU - Kamper, Herman

AU - Lopez, Adam

AU - Goldwater, Sharon

N1 - Conference code: 45

PY - 2020/5/14

Y1 - 2020/5/14

N2 - Given a large amount of unannotated speech in a low-resource language, can we classify the speech utterances by topic? We consider this question in the setting where a small amount of speech in the low-resource language is paired with text translations in a high-resource language. We develop an effective cross-lingual topic classifier by training on just 20 hours of translated speech, using a recent model for direct speech-to-text translation. While the translations are poor, they are still good enough to correctly classify the topic of 1-minute speech segments over 70% of the time—a 20% improvement over a majority-class baseline. Such a system could be useful for humanitarian applications like crisis response, where incoming speech in a foreign low-resource language must be quickly assessed for further action.

AB - Given a large amount of unannotated speech in a low-resource language, can we classify the speech utterances by topic? We consider this question in the setting where a small amount of speech in the low-resource language is paired with text translations in a high-resource language. We develop an effective cross-lingual topic classifier by training on just 20 hours of translated speech, using a recent model for direct speech-to-text translation. While the translations are poor, they are still good enough to correctly classify the topic of 1-minute speech segments over 70% of the time—a 20% improvement over a majority-class baseline. Such a system could be useful for humanitarian applications like crisis response, where incoming speech in a foreign low-resource language must be quickly assessed for further action.

KW - speech translation

KW - Low-resource speech processing

KW - speech classification

KW - unwritten languages

U2 - 10.1109/ICASSP40776.2020.9054169

DO - 10.1109/ICASSP40776.2020.9054169

M3 - Conference contribution

SN - 978-1-5090-6632-2

SP - 8164

EP - 8168

BT - ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

PB - Institute of Electrical and Electronics Engineers (IEEE)

T2 - 2020 IEEE International Conference on Acoustics, Speech, and Signal Processing

Y2 - 4 May 2020 through 8 May 2020

ER -

Cross-Lingual Topic Prediction for Speech Using Translations

Abstract

Publication series

Conference

Keywords

Access to Document

Fingerprint

Cite this