Datasetpublic

Khaledtelbahnasy/egyptian-arabic-stt-data

Egyptian Arabic STT Dataset Synthetic Egyptian Arabic speech dataset generated by the Synthetic Egyptian Speech Data Pipeline. Samples are human-reviewed and quality-validated using Whisper ASR (WER/CER). Dataset Statistics Metric Value Total samples 50 Total duration 85.2s (0.02h) Dialect validated 50 / 50 Average WER 0.4136 Average CER 0.1642 Topics food_ordering Fields Field Type Description id string… See the full description on the dataset page: https://huggingface.co/datasets/Khaledtelbahnasy/egyptian-arabic-stt-data.

sourceHugging Facecc-by-4.0updated 5mo agoView on Hugging Face

1likes16downloads

Dataset Card

Egyptian Arabic STT Dataset

Synthetic Egyptian Arabic speech dataset generated by the Synthetic Egyptian Speech Data Pipeline.

Samples are human-reviewed and quality-validated using Whisper ASR (WER/CER).

Dataset Statistics

Metric	Value
Total samples	50
Total duration	85.2s (0.02h)
Dialect validated	50 / 50
Average WER	0.4136
Average CER	0.1642
Topics	food_ordering

Fields

Field	Type	Description
`id`	string	Deterministic UUID5 per sample
`reference_text`	string	Original Egyptian Arabic prompt
`synthesis_text`	string	Text fed to TTS synthesis
`transcription_text`	string	Whisper ASR output
`normalized_reference_text`	string	ASR-normalised reference
`normalized_transcription_text`	string	ASR-normalised transcript
`wav_path`	string	Path to 16 kHz mono PCM WAV
`duration_seconds`	float	Audio duration
`wer`	float	Word Error Rate
`cer`	float	Character Error Rate
`dialect_validated`	bool	True unless Whisper detected non-Arabic ≥ 80%
`review_status`	string	Human decision: accepted / rejected / flagged
`reviewer_id`	string	Reviewer name
`provider`	string	TTS provider (nile\_tts / hamsa)
`topic`	string	Prompt topic category