metadata
dataset_info:
features:
- name: id
dtype: string
- name: audio
dtype: audio
- name: audio_language
dtype: string
- name: text
dtype: string
- name: transcription
dtype: string
- name: wer
dtype: float32
- name: prompt
dtype: string
- name: duration
dtype: float64
- name: speaker_id
dtype: string
splits:
- name: train
num_bytes: 5281189490
num_examples: 24869
download_size: 5274740145
dataset_size: 5281189490
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
language:
- zu
Transcriptions
The transcriptions in this dataset were generated using the TheirStory/whisper-medium-zulu model.
This model is a fine-tuned version of OpenAI's Whisper, specifically adapted for Zulu and related languages.