Darija ASR Dataset: Moroccan Arabic Dialect Speech Recordings

Moroccan Darija dialect speech recordings for automatic speech recognition. The dataset contains audio files in WAV format at 16 kHz mono, sourced from segmented YouTube excerpts. It was created by author mohamedmou and last updated on 2026-04-27.

Use Cases

• Train automatic speech recognition models based on the described Moroccan Darija audio.
• Fine-tune pre-trained ASR models for dialectal Arabic based on the described segmented YouTube audio.
• Benchmark ASR performance on non-standard Arabic varieties based on the described Darija language data.
• Study acoustic features of the Moroccan Arabic dialect based on the described 16 kHz mono recordings.

Strengths

• Audio is stored in a consistent, machine-readable format (WAV 16 kHz mono PCM 16-bit).
• Data is sourced from a diverse, real-world platform (YouTube excerpts).
• Dataset has a clear, defined task focus (Automatic Speech Recognition for Moroccan Darija).

Limitations

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

Provenance

Source

YouTube excerpts

Collection Method

Acoustically segmented extracts

Time Range

null

Freshness

Last updated 2026-04-27 18:20:01; freshness should be verified.

Geography

Morocco (implied by language focus)