GigaSpeech: An Evolving, Multi-domain ASR Corpus with 10, 000 Hours of Transcribed Audio

By Guoguo Chen, Shuzhou Chai, Guan-Bo Wang, Jiayu Du, Weiqiang Zhang, Chao Weng, Dan Su, Daniel Povey, Jan "Yenda" Trmal, Junbo Zhang and 11 more. 2021

607 citations recorded by the source corpus.

This paper introduces GigaSpeech, an evolving, multi-domain English speech recognition corpus with 10,000 hours of high quality labeled audio suitable for supervised training, and 40,000 hours of total audio suitable for semi-supervised and unsupervised training. Around 40,000 hours of transcribed audio is first collected from audiobooks, podcasts and YouTube, covering both read and spontaneous speaking styles, and a variety of topics, such as arts, science, sports, etc. A new forced alignment and segmentation pipeline is proposed to create sentence segments suitable for speech recognition training, and to filter out segments with low-quality transcription. For system training, GigaSpeech provides five subsets of different sizes, 10h, 250h, 1000h, 2500h, and 10000h. For our 10,000-hour XL training subset, we cap the word error rate at 4% during the filtering/validation stage, and for all our other smaller training subsets, we cap it at 0%. The DEV and TEST evaluation sets, on the other hand, are re-processed by professional human transcribers to ensure high transcription quality. Baseline systems are provided for popular speech recognition toolkits, namely Athena, ESPnet, Kaldi and Pika.

Subfields: Artificial Intelligence, Evaluation & Benchmarks, Speech & Audio, 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

In Evaluation & Benchmarks

In Speech & Audio

In Signal Processing

Citations in the corpus

Robust Speech Recognition via Large-Scale Weak Supervision (2022, neutral): “By relaxing the requirement of gold-standard human-validated transcripts, Chen et al. (2021) and Galvez et al. (2021) make use of sophisticated automated pipelines to scale weakly supervised speech recognition to 10,000 and 30,000 hours of noisier training data.”
The People's Speech: A Large-Scale Diverse English Speech Recognition Dataset for Commercial Usage (2021, background): “Gigaspeech [18] is a 10,000 hour English speech dataset.”
Ego4D: Around the World in 3,000 Hours of Egocentric Video (2021, background): “[34] Guoguo Chen, Shuzhou Chai, Guanbo Wang, Jiayu Du, WeiQiang Zhang, Chao Weng, Dan Su, Daniel Povey, Jan Trmal, Junbo Zhang, et al.”
Libriheavy: A 50,000 Hours ASR Corpus with Punctuation Casing and Context (2023, background): “In recent years, there are also large-scale corpora like GigaSpeech [9], People’s Speech [10] and MLS [11].”
FunCodec: A Fundamental, Reproducible and Integrable Open-Source Toolkit for Neural Speech Codec (2023, methodology): “To assess the generalization ability of our generalized model, we evaluate it on the test sets of multiple open-source corpora, including Librispeech [28], aishell-1 [29], aishell-2 [30], Wenet [31] and Gigaspeech [32].”
Mega-TTS: Zero-Shot Text-to-Speech at Scale with Intrinsic Inductive Bias (2023, methodology): “We use GigaSpeech [9] and WenetSpeech [65] as the training corpora, which contains 20k hours of multi-domain speeches in English and Chinese in total.”
Speak Foreign Languages with Your Own Voice: Cross-Lingual Neural Codec Language Modeling (2023, neutral): “Though researchers have put much effort into constructing speech-to-speech translation corpora, such as Voxvopule [Wang et al., 2021], CVSS [Jia et al., 2022b], and SpeechMatrix [Duquenne et al., 2022], they are either synthesized from text or mined from multilingual speech corpora thus can not…”
WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing (2021, methodology): “Thus, we just use the subset of 10k hours of GigaSpeech data, which is well processed and validated with a segmentation pipeline proposed in [16].”
Make-A-Voice: Revisiting Voice Large Language Models as Scalable Multilingual and Multitask Learners (2024, neutral): “• Librilight (Kahn et al., 2020) contains 60K hours of unlabelled speech from audiobooks in English, and LibriSpeech (Panayotov et al., 2015), Lib-riTTS (Zen et al., 2019), Gigaspeech (Chen et al., 2021), AISHELL (Fu et al., 2021), VCTK (Veaux et al., 2017) datasets include transcriptions.”
WENETSPEECH: A 10000+ Hours Multi-Domain Mandarin Corpus for Speech Recognition (2021, background): “Recently, a largescale multi-domain English corpus is released, which is named GigaSpeech [22].”
SpeechGPT: Empowering Large Language Models with Intrinsic Cross-Modal Conversational Abilities (2023, methodology): “For cross-modal instruction fine-tuning stage, we use Gigaspeech (Chen et al., 2021), Common voice (Ardila et al., 2020) and LibriSpeech (Panay-otov et al., 2015) dataset and moss-002-sft-data dataset, which is illustrated in detail in 3.1.”
Lhotse: a speech data representation library for the modern deep learning ecosystem (2021, methodology): “000 hours of recorded speech freely available with the recent releases of LibriSpeech [13], Multilingual LibriSpeech [15], LibriLight [9], GigaSpeech [7], SPGISpeech [12], The People’s Speech [8], and others.”
Self-Supervised Speech Representation Learning: A Review (2022, neutral): “Several datasets, including GigaSpeech [220], TED-LIUM 3 (TED3) [221], TED-LIUM 2 (TED2) [222], Switchboard (SWB) [223], TIMIT [96], and VoxLingua107 [224], are labeled”
LauraGPT: Listen, Attend, Understand, and Regenerate Audio with GPT (2023, methodology): “For the ASR task, we utilize open-source Chinese datasets such as AISHELL-1 (Bu et al., 2017), AISHELL-2 (Du et al., 2018), WenetSpeech (Zhang et al., 2022), as well as open-source English datasets including LibriSpeech (Panayotov et al., 2015) and GigaSpeech (Chen et al., 2021a).”
Leveraging Pseudo-labeled Data to Improve Direct Speech-to-Speech Translation (2022, methodology): “We utilize the in-house MT to convert ASR data that includes a subset of Gigaspeech [18] (Giga-Sub) and the Spanish subset of multilingual LibriSpeech [19] (MLS Es) to pseudo-labeled ST data, and then use the same in-house TTS to synthesize the target speech from the pseudo translation sequences.”
GigaST: A 10, 000-hour Pseudo Speech Translation Corpus (2022, methodology): “We start from GigaSpeech [7], a multi-domain English speech recognition corpus with 10,000 hours of labeled audio, and create paired text-to-text translation with an MT model.”