Xhosa Speech Corpus from the NCHLT Project

NCHLT Speech Corpus Xhosa contains audio recordings of the Xhosa language, a major South African language. The dataset was created by Beijuka and uploaded to Hugging Face in June 2024. It is part of the National Centre for Human Language Technology (NCHLT) initiative.

Use Cases

• Train automatic speech recognition models on Xhosa audio recordings.
• Benchmark speech-to-text systems using the provided transcriptions.
• Analyze phonetic features of Xhosa from the audio waveforms.
• Develop multilingual speech models by incorporating this corpus alongside other NCHLT datasets.

Strengths

• Part of the structured NCHLT project for South African languages.
• Available on Hugging Face with supporting libraries like datasets and mlcroissant.

Limitations

• Specific row count, audio duration, and speaker demographics are unknown.
• License terms and data collection methodology are not specified.
• Potential geographic bias limited to South African Xhosa speakers.

Provenance

Source

National Centre for Human Language Technology (NCHLT) project.

Freshness

Last updated on Hugging Face in June 2024.

Geography

South Africa (inferred from focus on Xhosa language).