TWB Voice 1.0: Multilingual Speech Corpus for Nigerian Languages

A multilingual speech corpus containing read speech data in three languages from Nigeria: Hausa, Shuwa Arabic, and Kanuri. It was created by CLEAR Global (formerly Translators without Borders) as part of the TWB Voice project to support automatic speech recognition development for underrepresented languages.

Use Cases

• Training automatic speech recognition (ASR) models based on the described read speech data.
• Benchmarking speech recognition performance for underrepresented languages based on the described Hausa, Shuwa Arabic, and Kanuri data.
• Developing language technology resources for Nigerian languages based on the multilingual speech corpus.
• Researching acoustic and linguistic features of the included languages based on the speech recordings.

Strengths

• Focuses on three underrepresented languages from Nigeria: Hausa, Shuwa Arabic, and Kanuri.
• Created by CLEAR Global, an organization with a stated mission to support language technology for underserved communities.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

CLEAR-Global

Collection Method

Likely recorded read speech as part of the TWB Voice project.

Freshness

Last updated 2025-08-07 10:00:03; freshness should be verified.

Geography

Nigeria