zionia/isixhosa-asr-specaugment

isiXhosa Speech Recognition Augmented Dataset - SpecAugment Dataset Description This dataset contains augmented speech recordings and transcriptions for isiXhosa, one of South Africa's official languages. The dataset has been optimized for use with OpenAI's Whisper ASR models. Dataset Statistics Number of samples: 686 Language: isiXhosa (Xho) Audio format: WAV, 16kHz, mono, 16-bit Maximum duration: 30 seconds (truncated for Whisper compatibility)… See the full description on the dataset page: https://huggingface.co/datasets/zionia/isixhosa-asr-specaugment.