A semi-automatic pipeline for transcribing and segmenting child speech

Polychronia  Christodoulidou; James Tanner; Jane Stuart-Smith; Michael McAuliffe; Mridhula Murali; Amy Smith; Lauren Taylor; Joanne Cleland; Anja Kuschmann

doi:10.21437/Interspeech.2025-1030

A semi-automatic pipeline for transcribing and segmenting child speech

Polychronia Christodoulidou, James Tanner, Jane Stuart-Smith, Michael McAuliffe, Mridhula Murali, Amy Smith, Lauren Taylor, Joanne Cleland, Anja Kuschmann

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

16 Downloads (Pure)

Abstract

This study evaluates both automated transcription (WhisperX) and forced alignment (MFA) in developing a semi-automated pipeline for obtaining acoustic vowel measures from field recordings from 275 children speaking a non-standard, English dialect, Scottish English. As expected, manual correction of speech transcriptions before forced alignment improves the quality of acoustic vowel measures with respect to manually-annotated data, though speech style and recording environment present some challenges for both tools. Adaptation of the MFA pre-trained english_us_arpa acoustic model towards the children's speech also improves the quality of acoustic measures, though greater improvement was not found by increasing training sample size.

Original language	English
Title of host publication	Proceedings of Interspeech 2025
Pages	4278-4282
Number of pages	5
DOIs	https://doi.org/10.21437/Interspeech.2025-1030
Publication status	Published - 14 Aug 2025
Event	Interspeech - Rotterdam, Rotterdam, Netherlands Duration: 17 Aug 2025 → 21 Aug 2025 https://www.interspeech2025.org/home

Publication series

Name	Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH
ISSN (Print)	2308-457X

Conference

Conference	Interspeech
Country/Territory	Netherlands
City	Rotterdam
Period	17/08/25 → 21/08/25
Internet address	https://www.interspeech2025.org/home

Funding

This research was supported by ESRC grant ES/W003244/1.

Keywords

child speech
automated speech processing
non-standard English
Whisper X
force alignment

Access to Document

10.21437/Interspeech.2025-1030

Christodoulidou-etal-Interspeech-2025-A-semi-automatic-pipeline-for-transcribing-and-segmenting-child-speechFinal published version, 545 KBLicence: Other

Cite this

Christodoulidou, P., Tanner, J., Stuart-Smith, J., McAuliffe, M., Murali, M., Smith, A., Taylor, L., Cleland, J., & Kuschmann, A. (2025). A semi-automatic pipeline for transcribing and segmenting child speech. In Proceedings of Interspeech 2025 (pp. 4278-4282). (Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH). https://doi.org/10.21437/Interspeech.2025-1030

@inproceedings{61737d8974d543a88538cb3bd7260f8b,

title = "A semi-automatic pipeline for transcribing and segmenting child speech",

abstract = "This study evaluates both automated transcription (WhisperX) and forced alignment (MFA) in developing a semi-automated pipeline for obtaining acoustic vowel measures from field recordings from 275 children speaking a non-standard, English dialect, Scottish English. As expected, manual correction of speech transcriptions before forced alignment improves the quality of acoustic vowel measures with respect to manually-annotated data, though speech style and recording environment present some challenges for both tools. Adaptation of the MFA pre-trained english\_us\_arpa acoustic model towards the children's speech also improves the quality of acoustic measures, though greater improvement was not found by increasing training sample size.",

keywords = "child speech, automated speech processing, non-standard English, Whisper X, force alignment",

author = "Polychronia Christodoulidou and James Tanner and Jane Stuart-Smith and Michael McAuliffe and Mridhula Murali and Amy Smith and Lauren Taylor and Joanne Cleland and Anja Kuschmann",

year = "2025",

month = aug,

day = "14",

doi = "10.21437/Interspeech.2025-1030",

language = "English",

series = "Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH",

pages = "4278--4282",

booktitle = "Proceedings of Interspeech 2025",

note = "Interspeech ; Conference date: 17-08-2025 Through 21-08-2025",

url = "https://www.interspeech2025.org/home",

}

Christodoulidou, P, Tanner, J, Stuart-Smith, J, McAuliffe, M, Murali, M , Smith, A, Taylor, L, Cleland, J & Kuschmann, A 2025, A semi-automatic pipeline for transcribing and segmenting child speech. in Proceedings of Interspeech 2025. Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH, pp. 4278-4282, Interspeech, Rotterdam, Netherlands, 17/08/25. https://doi.org/10.21437/Interspeech.2025-1030

A semi-automatic pipeline for transcribing and segmenting child speech. / Christodoulidou, Polychronia ; Tanner, James; Stuart-Smith, Jane et al.
Proceedings of Interspeech 2025. 2025. p. 4278-4282 (Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH).

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

TY - GEN

T1 - A semi-automatic pipeline for transcribing and segmenting child speech

AU - Christodoulidou, Polychronia

AU - Tanner, James

AU - Stuart-Smith, Jane

AU - McAuliffe, Michael

AU - Murali, Mridhula

AU - Smith, Amy

AU - Taylor, Lauren

AU - Cleland, Joanne

AU - Kuschmann, Anja

PY - 2025/8/14

Y1 - 2025/8/14

N2 - This study evaluates both automated transcription (WhisperX) and forced alignment (MFA) in developing a semi-automated pipeline for obtaining acoustic vowel measures from field recordings from 275 children speaking a non-standard, English dialect, Scottish English. As expected, manual correction of speech transcriptions before forced alignment improves the quality of acoustic vowel measures with respect to manually-annotated data, though speech style and recording environment present some challenges for both tools. Adaptation of the MFA pre-trained english_us_arpa acoustic model towards the children's speech also improves the quality of acoustic measures, though greater improvement was not found by increasing training sample size.

AB - This study evaluates both automated transcription (WhisperX) and forced alignment (MFA) in developing a semi-automated pipeline for obtaining acoustic vowel measures from field recordings from 275 children speaking a non-standard, English dialect, Scottish English. As expected, manual correction of speech transcriptions before forced alignment improves the quality of acoustic vowel measures with respect to manually-annotated data, though speech style and recording environment present some challenges for both tools. Adaptation of the MFA pre-trained english_us_arpa acoustic model towards the children's speech also improves the quality of acoustic measures, though greater improvement was not found by increasing training sample size.

KW - child speech

KW - automated speech processing

KW - non-standard English

KW - Whisper X

KW - force alignment

UR - https://www.interspeech2025.org/home

UR - https://www.isca-archive.org/interspeech_2025/christodoulidou25_interspeech.html

UR - https://www.scopus.com/pages/publications/105020072669

U2 - 10.21437/Interspeech.2025-1030

DO - 10.21437/Interspeech.2025-1030

M3 - Conference contribution book

T3 - Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH

SP - 4278

EP - 4282

BT - Proceedings of Interspeech 2025

T2 - Interspeech

Y2 - 17 August 2025 through 21 August 2025

ER -

A semi-automatic pipeline for transcribing and segmenting child speech

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Variability in child speech (VariCS)

Cite this

A semi-automatic pipeline for transcribing and segmenting child speech

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Projects

Variability in child speech (VariCS)

Cite this