Speech

Collecting and labelling audio for African languages. The tasks below share the same recording, consent and quality groundwork; they differ in what is produced from the audio.

Shared across speech data

• Speaker recruitment, consent & voice rights
• Recording setup & environment (device, microphone, background noise)
• Audio formats, sample rates & file standards
• Audio quality control (SNR, clipping, silence, channel checks)
• Transcription conventions (orthography, code-switching, disfluencies)
• Metadata (speaker demographics, device, environment, locale)
• Licensing & ethical handling of voices

Tasks

• ASR (Automatic Speech Recognition) – (transcription, multilingual ASR, code-switching)
• TTS (Text-to-Speech) – (single-speaker, multi-speaker, expressive TTS)
• Speech-to-Speech Translation (STS) – (direct speech translation across languages)
• Audio Understanding – (audio classification, sound event detection)
• Speech emotion recognition
• Speaker diarization