Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
The dataset viewer is not available for this split.
Server error while post-processing the rows. Please report the issue.
Error code:   RowsPostProcessingError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Dataset Description

Dialectra Yoruba Speech Corpus v1 is an open speech dataset collected from native Yoruba speakers across multiple regions and dialect communities.

The dataset was created to support the development of speech technologies for Yoruba, including Automatic Speech Recognition (ASR), Speech-to-Text (STT), Text-to-Speech (TTS), dialect research, and broader language technology applications.

This release contains transcribed speech recordings collected and verified through Dialectra's speech data platform.

Motivation

Despite being spoken by tens of millions of people across Nigeria and the global diaspora, Yoruba remains underrepresented in modern speech AI systems.

Dialectra was established to build dialect-aware speech infrastructure for African languages through:

  • Speech collection
  • Annotation
  • Human verification
  • Benchmarking
  • Dialect evaluation
  • Conversational speech infrastructure

This dataset is part of that mission.

Dataset Summary

Metric Value
Language Yoruba
Country Nigeria
Audio Format WAV
Sampling Rate 16 kHz
Dataset Version v1
Speakers 41
Audio Samples 310
Duration 40 minutes +
Collection Method Dialectra platform

Dialect Coverage

The dataset includes recordings from Yoruba speakers across different regions.

Examples include:

  • Standard Yoruba
  • Oyo
  • Ibadan
  • Ijebu
  • Ekiti
  • Ondo
  • Mixed Yoruba variants

Dialect representation may not be perfectly balanced.

Researchers are encouraged to evaluate model performance across dialect groups independently.

Data Collection Process

Speech samples were collected through the Dialectra platform.

Contributors voluntarily participated and recorded speech tasks using mobile and web interfaces.

Collection methods included:

  • Scripted speech
  • Prompt-based recordings
  • Read speech

All contributors agreed to the platform's contribution and licensing terms.

Intended Uses

Suitable For

  • Automatic Speech Recognition (ASR)
  • Speech-to-Text (STT)
  • Text-to-Speech (TTS)
  • Dialect Classification
  • Speech Research
  • Acoustic Modeling
  • Language Technology Development

Out-of-Scope Uses

  • Speaker Identification
  • Biometric Authentication
  • Demographic Profiling
  • Surveillance Applications

Benchmarking

Dialectra maintains internal benchmarking pipelines for evaluating speech quality and transcription accuracy.

Metrics may include:

  • Word Error Rate (WER)
  • Character Error Rate (CER)
  • Dialect Recognition Rate (DRR)
  • Dataset Coverage

Benchmark reports may be released separately.

Ethical Considerations

Researchers should consider:

  • Dialect representation imbalance
  • Regional variation
  • Demographic coverage
  • Potential biases within speech data

The dataset should be used responsibly and in compliance with applicable laws and regulations.

License

CC-BY-4.0

Citation

@dataset{dialectra_yoruba_2026, title={Dialectra Yoruba Speech Corpus v1}, author={Dialectra}, year={2026}, publisher={Hugging Face} }

About Dialectra

Dialectra is building dialect-aware speech infrastructure for African languages through speech collection, annotation, benchmarking, conversational datasets, and speech intelligence systems.

Website: https://dialectra.io

Contact: bakaka@dialectra.io 

Downloads last month
46