Segmenting Subtitles for Correcting ASR Segmentation Errors

David Wan; Chris Kedzie; Faisal Ladhak; Elsbeth Turcan; Petra Galuščáková; Elena Zotkina; Zhengping Jiang; Peter Bell; Kathleen McKeown

Segmenting Subtitles for Correcting ASR Segmentation Errors

David Wan, Chris Kedzie, Faisal Ladhak, Elsbeth Turcan, Petra Galuščáková, Elena Zotkina, Zhengping Jiang, Peter Bell, Kathleen McKeown

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

Abstract

Typical ASR systems segment the input audio in toutterances using purely acoustic information, which may not resemble the sentence-like units that are expected by conventional machine translation (MT) systems for Spoken Language Translation. In this work, we propose a model for correcting the acoustic segmentation of ASR models for low-resource languages to improve performance on down-stream tasks. We propose the use of subtitles as a proxy dataset for correcting ASR acoustic segmentation, creating synthetic acoustic utterances by modeling common error modes. We train a neural tagging model for correcting ASR acoustic segmentation and show that it improves downstream performance on MT and audio-document cross-language information retrieval (CLIR).

Original language	English
Title of host publication	Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume
Publisher	Association for Computational Linguistics
Pages	2842-2854
Number of pages	13
ISBN (Print)	978-1-954085-02-2
Publication status	Published - 19 Apr 2021
Event	16th conference of the European Chapter of the Association for Computational Linguistics - Virtual Conference Duration: 19 Apr 2021 → 23 Apr 2021 https://2021.eacl.org/

Conference

Conference	16th conference of the European Chapter of the Association for Computational Linguistics
Abbreviated title	EACL 2021
City	Virtual Conference
Period	19/04/21 → 23/04/21
Internet address	https://2021.eacl.org/

Access to Document

Segmenting Subtitles_WAN_DOA11012021_VOR_CC-BYFinal published version, 377 KBLicence: Creative Commons: Attribution (CC-BY)

https://www.aclweb.org/anthology/2021.eacl-main.248Licence: Creative Commons: Attribution (CC-BY)

Cite this

Wan, D., Kedzie, C., Ladhak, F., Turcan, E., Galuščáková, P., Zotkina, E., Jiang, Z., Bell, P., & McKeown, K. (2021). Segmenting Subtitles for Correcting ASR Segmentation Errors. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume (pp. 2842-2854). Association for Computational Linguistics. https://www.aclweb.org/anthology/2021.eacl-main.248

@inproceedings{69152d1a52014827aaca683c6f9862e0,

title = "Segmenting Subtitles for Correcting ASR Segmentation Errors",

abstract = "Typical ASR systems segment the input audio in toutterances using purely acoustic information, which may not resemble the sentence-like units that are expected by conventional machine translation (MT) systems for Spoken Language Translation. In this work, we propose a model for correcting the acoustic segmentation of ASR models for low-resource languages to improve performance on down-stream tasks. We propose the use of subtitles as a proxy dataset for correcting ASR acoustic segmentation, creating synthetic acoustic utterances by modeling common error modes. We train a neural tagging model for correcting ASR acoustic segmentation and show that it improves downstream performance on MT and audio-document cross-language information retrieval (CLIR).",

author = "David Wan and Chris Kedzie and Faisal Ladhak and Elsbeth Turcan and Petra Galu{\v s}{\v c}{\'a}kov{\'a} and Elena Zotkina and Zhengping Jiang and Peter Bell and Kathleen McKeown",

year = "2021",

month = apr,

day = "19",

language = "English",

isbn = "978-1-954085-02-2",

pages = "2842--2854",

booktitle = "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume",

publisher = "Association for Computational Linguistics",

note = "16th conference of the European Chapter of the Association for Computational Linguistics , EACL 2021 ; Conference date: 19-04-2021 Through 23-04-2021",

url = "https://2021.eacl.org/",

}

Wan, D, Kedzie, C, Ladhak, F, Turcan, E, Galuščáková, P, Zotkina, E, Jiang, Z, Bell, P & McKeown, K 2021, Segmenting Subtitles for Correcting ASR Segmentation Errors. in Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume. Association for Computational Linguistics, pp. 2842-2854, 16th conference of the European Chapter of the Association for Computational Linguistics , Virtual Conference, 19/04/21. <https://www.aclweb.org/anthology/2021.eacl-main.248>

Segmenting Subtitles for Correcting ASR Segmentation Errors. / Wan, David; Kedzie, Chris; Ladhak, Faisal et al.

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume. Association for Computational Linguistics, 2021. p. 2842-2854.

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

TY - GEN

T1 - Segmenting Subtitles for Correcting ASR Segmentation Errors

AU - Wan, David

AU - Kedzie, Chris

AU - Ladhak, Faisal

AU - Turcan, Elsbeth

AU - Galuščáková, Petra

AU - Zotkina, Elena

AU - Jiang, Zhengping

AU - Bell, Peter

AU - McKeown, Kathleen

PY - 2021/4/19

Y1 - 2021/4/19

N2 - Typical ASR systems segment the input audio in toutterances using purely acoustic information, which may not resemble the sentence-like units that are expected by conventional machine translation (MT) systems for Spoken Language Translation. In this work, we propose a model for correcting the acoustic segmentation of ASR models for low-resource languages to improve performance on down-stream tasks. We propose the use of subtitles as a proxy dataset for correcting ASR acoustic segmentation, creating synthetic acoustic utterances by modeling common error modes. We train a neural tagging model for correcting ASR acoustic segmentation and show that it improves downstream performance on MT and audio-document cross-language information retrieval (CLIR).

AB - Typical ASR systems segment the input audio in toutterances using purely acoustic information, which may not resemble the sentence-like units that are expected by conventional machine translation (MT) systems for Spoken Language Translation. In this work, we propose a model for correcting the acoustic segmentation of ASR models for low-resource languages to improve performance on down-stream tasks. We propose the use of subtitles as a proxy dataset for correcting ASR acoustic segmentation, creating synthetic acoustic utterances by modeling common error modes. We train a neural tagging model for correcting ASR acoustic segmentation and show that it improves downstream performance on MT and audio-document cross-language information retrieval (CLIR).

M3 - Conference contribution

SN - 978-1-954085-02-2

SP - 2842

EP - 2854

BT - Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume

PB - Association for Computational Linguistics

T2 - 16th conference of the European Chapter of the Association for Computational Linguistics

Y2 - 19 April 2021 through 23 April 2021

ER -

Segmenting Subtitles for Correcting ASR Segmentation Errors

Abstract

Conference

Access to Document

Fingerprint

Cite this