metadata
dataset_info:
features:
- name: client_id
dtype: string
- name: audio
dtype:
audio:
sampling_rate: 16000
- name: sentence
dtype: string
- name: up_votes
dtype: int64
- name: down_votes
dtype: int64
- name: age
dtype: string
- name: gender
dtype: string
- name: accent
dtype: string
- name: locale
dtype: string
- name: segment
dtype: string
- name: variant
dtype: string
- name: n_letters
dtype: int64
- name: path
dtype: string
splits:
- name: train
num_bytes: 3293820
num_examples: 100
download_size: 3271366
dataset_size: 3293820
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
This dataset is a subset of Mozilla Common Voice 16.1, focusing on Swahili training speaker with ID# 7ec265470aa1c06d5b156e0c67c5014cfaf580e7074798f11787f82cbc3fe0ef6aefe2b969e827a41486dc4f0c034b7444ad184067378600ad5f2b9656e2a929.
It has been filtered to the first 100 rows which are higher quality audio.
The equivalent dataset subset can be accessed here.