Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

WAXAL: A Large-Scale Multilingual African Language Speech Corpus

Abdoulaye Diack, Perry Nelson, Kwaku Agbesi, Angela Nakalembe, MohamedElfatih MohamedKhair, Vusumuzi Dube, Tavonga Siyavora, Subhashini Venugopalan, Jason Hickey, Uche Okonkwo, Abhishek Bapna, Isaac Wiafe, Raynard Dodzi Helegah, Elikem Doe Atsakpo, Charles Nutrokpor, Fiifi Baffoe Payin Winful, Kafui Kwashie Solaga, Jamal-Deen Abdulai, Akon Obu Ekpezu, Audace Niyonkuru, Samuel Rutunda, Boris Ishimwe, Michael Melese, Engineer Bainomugisha, Joyce Nakatumba-Nabende, Andrew Katumba, Claire Babirye, Jonathan Mukiibi, Vincent Kimani, Samuel Kibacia, James Maina, Fridah Emmah, Ahmed Ibrahim Shekarau, Ibrahim Shehu Adamu, Yusuf Abdullahi, Howard Lakougna, Bob MacDonald, Hadar Shemtov, Aisha Walcott-Bryant, Moustapha Cisse, Avinatan Hassidim, Jeff Dean, Yossi Matias

TL;DR

The paper tackles the data scarcity barrier hindering speech technology for Sub-Saharan African languages by introducing WAXAL, a large-scale multilingual corpus covering 21 languages. It describes a multi-year, partner-driven data collection effort that yields ~1,250 hours of transcribed ASR data across 14 languages and ~186 hours of high-quality, single-speaker TTS data across 10 languages, with image-prompted ASR and phonetically balanced TTS scripts. The dataset is accompanied by rich metadata, local linguistic transcription, and quality-control processes, and is released under CC-BY-4.0 to maximize impact. The authors also discuss limitations (e.g., partial transcription coverage and dialectal variation) and ethical considerations (consent, privacy, and speaker rights), framing WAXAL as a foundation for inclusive, digitally preserved African languages and enabling broader research and development of accessible speech technologies.

Abstract

The advancement of speech technology has predominantly favored high-resource languages, creating a significant digital divide for speakers of most Sub-Saharan African languages. To address this gap, we introduce WAXAL, a large-scale, openly accessible speech dataset for 21 languages representing over 100 million speakers. The collection consists of two main components: an Automated Speech Recognition (ASR) dataset containing approximately 1,250 hours of transcribed, natural speech from a diverse range of speakers, and a Text-to-Speech (TTS) dataset with over 180 hours of high-quality, single-speaker recordings reading phonetically balanced scripts. This paper details our methodology for data collection, annotation, and quality control, which involved partnerships with four African academic and community organizations. We provide a detailed statistical overview of the dataset and discuss its potential limitations and ethical considerations. The WAXAL datasets are released at https://huggingface.co/datasets/google/WaxalNLP under the permissive CC-BY-4.0 license to catalyze research, enable the development of inclusive technologies, and serve as a vital resource for the digital preservation of these languages.

WAXAL: A Large-Scale Multilingual African Language Speech Corpus

TL;DR

The paper tackles the data scarcity barrier hindering speech technology for Sub-Saharan African languages by introducing WAXAL, a large-scale multilingual corpus covering 21 languages. It describes a multi-year, partner-driven data collection effort that yields ~1,250 hours of transcribed ASR data across 14 languages and ~186 hours of high-quality, single-speaker TTS data across 10 languages, with image-prompted ASR and phonetically balanced TTS scripts. The dataset is accompanied by rich metadata, local linguistic transcription, and quality-control processes, and is released under CC-BY-4.0 to maximize impact. The authors also discuss limitations (e.g., partial transcription coverage and dialectal variation) and ethical considerations (consent, privacy, and speaker rights), framing WAXAL as a foundation for inclusive, digitally preserved African languages and enabling broader research and development of accessible speech technologies.

Abstract

The advancement of speech technology has predominantly favored high-resource languages, creating a significant digital divide for speakers of most Sub-Saharan African languages. To address this gap, we introduce WAXAL, a large-scale, openly accessible speech dataset for 21 languages representing over 100 million speakers. The collection consists of two main components: an Automated Speech Recognition (ASR) dataset containing approximately 1,250 hours of transcribed, natural speech from a diverse range of speakers, and a Text-to-Speech (TTS) dataset with over 180 hours of high-quality, single-speaker recordings reading phonetically balanced scripts. This paper details our methodology for data collection, annotation, and quality control, which involved partnerships with four African academic and community organizations. We provide a detailed statistical overview of the dataset and discuss its potential limitations and ethical considerations. The WAXAL datasets are released at https://huggingface.co/datasets/google/WaxalNLP under the permissive CC-BY-4.0 license to catalyze research, enable the development of inclusive technologies, and serve as a vital resource for the digital preservation of these languages.
Paper Structure (10 sections, 4 figures, 1 table)

This paper contains 10 sections, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Works
  3. Data Collection Methodology
  4. ASR Data Collection and Annotation
  5. TTS Data Collection
  6. Dataset Statistics
  7. Limitations and Ethical Considerations
  8. Limitations
  9. Ethical Considerations
  10. Conclusion

Figures (4)

  • Figure 1: Examples of some of the images used as prompts to elicit natural speech for the ASR dataset.
  • Figure 2: WAXAL-ASR: Distribution of the ASR dataset sliced by language.
  • Figure 3: WAXAL-ASR: Dist. of age and gender of speakers.
  • Figure 4: WAXAL-TTS: Distribution of the TTS dataset sliced by language.