AfriSpeech/grn-african-speech

Domain:

natural language processing

Record type:

dataset

Creator:

Afr

Host:

15-second, 16kHz mono voice-only clips extracted from Global Recordings Network language recordings for African countries. Clips containing music, singing or background noise were removed with a YAMNet-based strict filter; only clean speech is kept. Audio (c) Global Recordings Network, used for research purposes. Please consult GRN's terms for any other use. Subsets: 1551 | Total speech: 3077.9 hours subset language clips

Visit

huggingface.co

Tasks

speech processing