michsethowusu/jw-african-speech
Domain:
natural language processingRecord type:
dataset
Creator:
mic
Host:
Paragraph-level audio/text pairs for African languages, built from
publicly available magazine audio recordings and article text on jw.org.
Clips are cut at the publisher's own paragraph timing markers and
filtered to 3-30 s. Languages whose audio has no timing markers are
not included.
Copyright notice: all audio and text content belongs to the
Watch Tower Bible and Tract Society of Pennsylvania (jw.org).