Libri-Light: A Benchmark for ASR with Limited or No Supervision
Jacob Kahn, Morgane Rivière, Weiyi Zheng, Evgeny Kharitonov, Qiantong Xu, Pierre-Emmanuel Mazaré, Julien Karadayi, Vitaliy Liptchinsky, Ronan Collobert, Christian Fuegen, Tatiana Likhomanenko, Gabriel Synnaeve, Armand Joulin, Abdel-rahman Mohamed, Emmanuel Dupoux
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/facebookresearch/libri-lightOfficialIn paperpytorch★ 0
- github.com/k2-fsa/libriheavynone★ 216
Abstract
We introduce a new collection of spoken English audio suitable for training speech recognition systems under limited or no supervision. It is derived from open-source audio books from the LibriVox project. It contains over 60K hours of audio, which is, to our knowledge, the largest freely-available corpus of speech. The audio has been segmented using voice activity detection and is tagged with SNR, speaker ID and genre descriptions. Additionally, we provide baseline systems and evaluation metrics working under three settings: (1) the zero resource/unsupervised setting (ABX), (2) the semi-supervised setting (PER, CER) and (3) the distant supervision setting (WER). Settings (2) and (3) use limited textual resources (10 minutes to 10 hours) aligned with the speech. Setting (3) uses large amounts of unaligned text. They are evaluated on the standard LibriSpeech dev and test sets for comparison with the supervised state-of-the-art.
Tasks
Benchmark Results
| Dataset | Model | Metric | Claimed | Verified | Status |
|---|---|---|---|---|---|
| Libri-Light test-clean | TDS 60k pseudo-label + CTC fine-tuning + 4gram-LM | Word Error Rate (WER) | 29.3 | — | Unverified |
| Libri-Light test-clean | CPC unlab-60k+train-10h CPC pretrain + CTC fine-tuning + 4gram-LM | Word Error Rate (WER) | 43.9 | — | Unverified |
| Libri-Light test-clean | CPC unlab-60k | ABX-across | 7.56 | — | Unverified |
| Libri-Light test-other | TDS 60k pseudo-label + CTC fine-tuning + 4gram-LM | Word Error Rate (WER) | 56.6 | — | Unverified |
| Libri-Light test-other | CPC unlab-60k+train-10h CPC pretrain + CTC fine-tuning + 4gram-LM | Word Error Rate (WER) | 69.5 | — | Unverified |
| Libri-Light test-other | CPC unlab-60k | ABX-across | 13.42 | — | Unverified |