Egyptian-ASR-MGB-3: Egyptian Arabic Dialect Speech for ASR Training

16 hours of Egyptian Arabic dialect speech audio manually transcribed for automatic speech recognition tasks. The dataset was collected from multi-genre YouTube channels, cleaned, and adjusted for the Hugging Face Hub by MightyStudent. It is intended for fine-tuning or training models like Whisper.

Use Cases

• Fine-tuning Whisper models for Egyptian Arabic speech recognition based on the described audio-transcript pairs.
• Training new automatic speech recognition models on manually transcribed dialectal audio.
• Benchmarking ASR performance on multi-genre, real-world Egyptian Arabic speech data.
• Studying linguistic features and variations within the Egyptian Arabic dialect.

Strengths

• 16 hours of manually transcribed audio data.
• Focus on a specific Arabic dialect (Egyptian) with real-world multi-genre sources.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.
• Data may reflect geographic and source bias inherent to its collection from specific YouTube channels.

Provenance

Source

MGB-3 website, via YouTube channels.

Collection Method

Collected, cleaned, and adjusted for the Hugging Face Hub.

Freshness

Last updated 2024-09-03 19:58:11; freshness should be verified.

Geography

Egyptian Arabic dialect.