Librispeech: An ASR corpus based on public domain audio books
By Vassil Panayotov, Guoguo Chen, Daniel Povey, Sanjeev Khudanpur. 2015
8,192 citations recorded by the source corpus.
This paper introduces a new corpus of read English speech, suitable for training and evaluating speech recognition systems. The LibriSpeech corpus is derived from audiobooks that are part of the LibriVox project, and contains 1000 hours of speech sampled at 16 kHz. We have made the corpus freely available for download, along with separately prepared language-model training data and pre-built language models. We show that acoustic models trained on LibriSpeech give lower error rate on the Wall Street Journal (WSJ) test sets than models trained on WSJ itself. We are also releasing Kaldi scripts that make it easy to build these systems.
Subfields: Artificial Intelligence, Evaluation & Benchmarks, Signal Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Yoshua Bengio
- Mohit Bansal
- Quoc V. Le
- Jianfeng Gao
- Percy Liang
- Luke S. Zettlemoyer
- Graham Neubig
- James R. Glass
In Evaluation & Benchmarks
- Yoshua Bengio
- James R. Glass
- Shinji Watanabe
- Luke S. Zettlemoyer
- Yu Zhang
- Mohit Bansal
- Yonatan Belinkov
- Graham Neubig
In Signal Processing
Citations in the corpus
- Representation Learning with Contrastive Predictive Coding (2018, methodology): “tions, and in reinforcement learning we measure how the auxiliary CPC loss speeds up learning of the agent. 3.1 Audio For audio, we use a 100-hour subset of the publicly available LibriSpeech dataset [30]. Although the dataset does not provide labels other than the raw text, we obtained force-aligne”
- Neural Discrete Representation Learning (2017, methodology): “For this task we have used a larger dataset of 460 speakers [30] and trained a VQ-VAE model where the resolution of discrete space is 128 times smaller.”
- Imputer: Sequence Modelling via Imputation and Dynamic Programming (2020, methodology): “We experiment with two competitive speech tasks, the 82 hours Wall Street Journal (WSJ) (Paul & Baker, 1992) dataset and the 960 hours LibriSpeech (Panayotov et al., 2015) dataset.”
- End-to-end ASR: from Supervised to Semi-Supervised Learning with Modern Architectures (2019, positive): “We set out to study end-to-end systems on LIBRISPEECH (Panayotov et al., 2015) and, without any algorithmic contribution, see if they can be made to perform as well as more complex training pipelines.”
- RWTH ASR Systems for LibriSpeech: Hybrid vs Attention - w/o Data Augmentation (2019, methodology): “We use the 4-gram count model officially distributed with the LibriSpeech dataset [2].”
- Robust Speech Recognition via Large-Scale Weak Supervision (2022, methodology): “• LibriSpeech (Panayotov et al., 2015): We used the test-clean and test-other splits from the LibriSpeech ASR corpus.”
- FLEURS: FEW-Shot Learning Evaluation of Universal Representations of Speech (2022, background): “0 [5] have demonstrated strong performance on the multilingual LibriSpeech dataset [10], in particular in the few-shot learning scenario with only 10 minutes of labeled data [11].”
- XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale (2021, background): “LibriSpeech is a widely-used evaluation benchmark for speech recognition research (Panayotov et al., 2015).”
- The People's Speech: A Large-Scale Diverse English Speech Recognition Dataset for Commercial Usage (2021, methodology): “This work is a first step towards creating a speech dataset that covers all speakers and environments (not just English audiobooks [22, 25, 26]) for open use [27].”
- Unispeech-Sat: Universal Speech Representation Learning With Speaker Aware Pre-Training (2021, positive): “We pretrain the UniSpeech-SAT Base model for 400k steps on LibriSpeech 960 hours audio [20] using the label generated by clustering the 6-th trans-former layer output of the first iteration model of HuBERT Base model.”
- BigSSL: Exploring the Frontier of Large-Scale Semi-Supervised Learning for Automatic Speech Recognition (2021, methodology): “By comparing the second and third rows, we see that pre-training on YT-U is able to benefit a wider variety of domains compared to LibriSpeech.”
- HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units (2021, methodology): “When the HuBERT model is pre-trained on either the standard Librispeech 960h [24] or the Libri-Light 60k hours [25], it either matches or improves upon the state-of-theart wav2vec 2.”
- GigaSpeech: An Evolving, Multi-domain ASR Corpus with 10, 000 Hours of Transcribed Audio (2021, positive): “Even LibriSpeech [6], one of the most popular corpora for speech recognition tasks, is more than 5 years old, and has a WER of 1.”
- Unsupervised Speech Recognition (2021, background): “Introduction Speech recognition performance on the much studied English Librispeech benchmark (Panayotov et al., 2015) has seen rapid improvement over the last few years due to advances in model architectures (Dong et al.”
- Earnings-21: A Practical Benchmark for ASR in the Wild (2021, methodology): “For further WER comparison, we also used an open-source Kaldi model(8) trained on 960 hours from LibriSpeech [4] with standard 3-way speed perturbation.”
- Robust wav2vec 2.0: Analyzing Domain Shift in Self-Supervised Pre-Training (2021, methodology): “The six domains are: (1) LibriSpeech [26] ( LS ) and Libri-light [27] ( LL ), which contain about 960 and 60K hours of 16kHz crowd-sourced audiobook recordings, respectively, derived from the LibriVox project; (2) TED-LIUM v3 [28] ( TD ), which contains 452 hours of TED conference audio in 16kHz; (3”
- VoxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretation (2021, methodology): “This line of research leverages large amounts of unlabelled English speech data (Kahn et al., 2020b) that enable improvements in English ASR or out-of-English ST. Large amounts of multilingual audio data are needed in order to achieve similar progress for multilingual ASR and ST. Similarly, most ASR”