A Cookbook for Community-driven Data Collection of Impaired Speech in Low-Resource Languages

Sumaya Ahmed Salihs; Isaac Wiafe; Jamal Deen Abdulai; Elikem Doe Atsakpo; Gifty Ayoka; Richard Cave; Akon Obu Ekpezu; Catherine Holloway; Katrin Tomanek; Fiifi Baffoe Payin Winful

doi:10.21437/Interspeech.2025-2261

A Cookbook for Community-driven Data Collection of Impaired Speech in Low-Resource Languages

Sumaya Ahmed Salihs, Isaac Wiafe, Jamal Deen Abdulai, Elikem Doe Atsakpo, Gifty Ayoka, Richard Cave, Akon Obu Ekpezu, Catherine Holloway, Katrin Tomanek, Fiifi Baffoe Payin Winful

Department of Computer Science

Research output: Contribution to journal › Conference article › peer-review

Abstract

This study presents an approach for collecting speech samples to build Automatic Speech Recognition (ASR) models for impaired speech, particularly, low-resource languages. It aims to democratize ASR technology and data collection by developing a "cookbook" of best practices and training for community-driven data collection and ASR model building. As a proof-of-concept, this study curated the first open-source dataset of impaired speech in Akan: a widely spoken indigenous language in Ghana. The study involved participants from diverse backgrounds with speech impairments. The resulting dataset, along with the cookbook and open-source tools, are publicly available to enable researchers and practitioners to create inclusive ASR technologies tailored to the unique needs of speech impaired individuals. In addition, this study presents the initial results of finetuning open-source ASR models to better recognize impaired speech in Akan.

Original language	English
Pages (from-to)	4623-4627
Number of pages	5
Journal	Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH
DOIs	https://doi.org/10.21437/Interspeech.2025-2261
Publication status	Published - 2025
Event	26th Interspeech Conference 2025 - Rotterdam Duration: 17 Aug 2025 → 21 Aug 2025

Keywords

automatic speech recognition
community engagement
democratizing AI
impaired speech
low resource language

Access to Document

10.21437/Interspeech.2025-2261

Cite this

Salihs, S. A., Wiafe, I., Abdulai, J. D., Atsakpo, E. D., Ayoka, G., Cave, R., Ekpezu, A. O., Holloway, C., Tomanek, K., & Winful, F. B. P. (2025). A Cookbook for Community-driven Data Collection of Impaired Speech in Low-Resource Languages. Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH, 4623-4627. https://doi.org/10.21437/Interspeech.2025-2261

@article{c65203f0d3ef427cad9fe136d6d824ac,

title = "A Cookbook for Community-driven Data Collection of Impaired Speech in Low-Resource Languages",

abstract = "This study presents an approach for collecting speech samples to build Automatic Speech Recognition (ASR) models for impaired speech, particularly, low-resource languages. It aims to democratize ASR technology and data collection by developing a {"}cookbook{"} of best practices and training for community-driven data collection and ASR model building. As a proof-of-concept, this study curated the first open-source dataset of impaired speech in Akan: a widely spoken indigenous language in Ghana. The study involved participants from diverse backgrounds with speech impairments. The resulting dataset, along with the cookbook and open-source tools, are publicly available to enable researchers and practitioners to create inclusive ASR technologies tailored to the unique needs of speech impaired individuals. In addition, this study presents the initial results of finetuning open-source ASR models to better recognize impaired speech in Akan.",

keywords = "automatic speech recognition, community engagement, democratizing AI, impaired speech, low resource language",

author = "Salihs, \{Sumaya Ahmed\} and Isaac Wiafe and Abdulai, \{Jamal Deen\} and Atsakpo, \{Elikem Doe\} and Gifty Ayoka and Richard Cave and Ekpezu, \{Akon Obu\} and Catherine Holloway and Katrin Tomanek and Winful, \{Fiifi Baffoe Payin\}",

note = "Publisher Copyright: {\textcopyright} 2025 International Speech Communication Association. All rights reserved.; 26th Interspeech Conference 2025 ; Conference date: 17-08-2025 Through 21-08-2025",

year = "2025",

doi = "10.21437/Interspeech.2025-2261",

language = "English",

pages = "4623--4627",

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

issn = "2308-457X",

publisher = "International Speech Communication Association",

}

Salihs, SA, Wiafe, I, Abdulai, JD, Atsakpo, ED, Ayoka, G, Cave, R, Ekpezu, AO, Holloway, C, Tomanek, K & Winful, FBP 2025, 'A Cookbook for Community-driven Data Collection of Impaired Speech in Low-Resource Languages', Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH, pp. 4623-4627. https://doi.org/10.21437/Interspeech.2025-2261

TY - JOUR

T1 - A Cookbook for Community-driven Data Collection of Impaired Speech in Low-Resource Languages

AU - Salihs, Sumaya Ahmed

AU - Wiafe, Isaac

AU - Abdulai, Jamal Deen

AU - Atsakpo, Elikem Doe

AU - Ayoka, Gifty

AU - Cave, Richard

AU - Ekpezu, Akon Obu

AU - Holloway, Catherine

AU - Tomanek, Katrin

AU - Winful, Fiifi Baffoe Payin

PY - 2025

Y1 - 2025

N2 - This study presents an approach for collecting speech samples to build Automatic Speech Recognition (ASR) models for impaired speech, particularly, low-resource languages. It aims to democratize ASR technology and data collection by developing a "cookbook" of best practices and training for community-driven data collection and ASR model building. As a proof-of-concept, this study curated the first open-source dataset of impaired speech in Akan: a widely spoken indigenous language in Ghana. The study involved participants from diverse backgrounds with speech impairments. The resulting dataset, along with the cookbook and open-source tools, are publicly available to enable researchers and practitioners to create inclusive ASR technologies tailored to the unique needs of speech impaired individuals. In addition, this study presents the initial results of finetuning open-source ASR models to better recognize impaired speech in Akan.

AB - This study presents an approach for collecting speech samples to build Automatic Speech Recognition (ASR) models for impaired speech, particularly, low-resource languages. It aims to democratize ASR technology and data collection by developing a "cookbook" of best practices and training for community-driven data collection and ASR model building. As a proof-of-concept, this study curated the first open-source dataset of impaired speech in Akan: a widely spoken indigenous language in Ghana. The study involved participants from diverse backgrounds with speech impairments. The resulting dataset, along with the cookbook and open-source tools, are publicly available to enable researchers and practitioners to create inclusive ASR technologies tailored to the unique needs of speech impaired individuals. In addition, this study presents the initial results of finetuning open-source ASR models to better recognize impaired speech in Akan.

KW - automatic speech recognition

KW - community engagement

KW - democratizing AI

KW - impaired speech

KW - low resource language

UR - https://www.scopus.com/pages/publications/105020037127

U2 - 10.21437/Interspeech.2025-2261

DO - 10.21437/Interspeech.2025-2261

M3 - Conference article

AN - SCOPUS:105020037127

SN - 2308-457X

SP - 4623

EP - 4627

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

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

T2 - 26th Interspeech Conference 2025

Y2 - 17 August 2025 through 21 August 2025

ER -

A Cookbook for Community-driven Data Collection of Impaired Speech in Low-Resource Languages

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this