Librispeech: 1000 hours of read English speech with sampling rate of 16 kHz.
train split: "train-clean-100" has 28539 examples (251 speakers)
- speaker_id
- chapter id
- speaker_id_chapter_id_000.flac
- speaker_id_chapter_id_000.flac
- ...
- speaker_id_chapter_id.trans.txt
- speaker_id
...
test split: "test-clean" has 2620 examples. (40 speakers)
$ python3 experiment.py
$ python3 transcribe.py