Common Voice is a large multilingual speech corpus with sentence-level transcripts. Adding a loader would enable training on languages beyond English.
Scope:
- Parse Common Voice TSV metadata (
validated.tsv)
- Load MP3 audio files via symphonia
- Map to
TrainingSample format
- Handle the flat
clips/ directory structure
Common Voice is a large multilingual speech corpus with sentence-level transcripts. Adding a loader would enable training on languages beyond English.
Scope:
validated.tsv)TrainingSampleformatclips/directory structure