deepdml/openslr-32-hq-SA-languages

SLR32 – High Quality TTS Data for Four South African Languages

Identifier: SLR32
License: CC BY-SA 4.0
Source: https://www.openslr.org/32/

This dataset contains multi-speaker high quality transcribed audio data for four languages of South Africa: Afrikaans (af_za), Sesotho (st_za), Setswana (tn_za) and isiXhosa (xh_za).

The dataset consists of WAV files and a TSV file transcribing the audio. In each folder the file line_index.tsv contains a FileID (which in turn encodes the UserID) and the transcription of the corresponding audio file.

The dataset has had some quality checks, but there might still be errors.

It was collected as a collaboration between North West University and Google.

Configs / Splits

Config	Language	Utterances
af_za	Afrikaans	2927
st_za	Sesotho	2096
tn_za	Setswana	2378
xh_za	isiXhosa	2420

Fields

Field	Type	Description
file_id	string	Unique identifier (encodes speaker/user)
audio	Audio	WAV audio
transcript	string	Transcription of the utterance
duration	float32	Audio duration in seconds

Citation

@inproceedings{van-niekerk-etal-2017,
  title     = {{Rapid development of TTS corpora for four South African languages}},
  author    = {Daniel van Niekerk and Charl van Heerden and Marelie Davel and
               Neil Kleynhans and Oddur Kjartansson and Martin Jansche and Linne Ha},
  booktitle = {Proc. Interspeech 2017},
  pages     = {2178--2182},
  address   = {Stockholm, Sweden},
  month     = aug,
  year      = {2017},
  URL       = {http://dx.doi.org/10.21437/Interspeech.2017-1139}
}

Copyright 2017 Google, Inc.