{
  "name": "Canonical Speech Recognition Benchmark Datasets Comparison",
  "description": "Comparative specifications of canonical automatic speech recognition corpora, detailing audio duration, sampling rates, speaker counts, acoustic domains, licensing terms, and primary research utility, compiled from official repository documentations and peer-reviewed conference publications.",
  "page": "https://casrai.org/guides/librispeech",
  "publisher": "CASRAI",
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "dateModified": "2026-11-16T18:59:32Z",
  "temporalCoverage": "1993/2026",
  "spatialCoverage": "Global",
  "columns": [
    {
      "name": "Corpus Name",
      "unit": "",
      "description": "Canonical dataset title and standard identifier"
    },
    {
      "name": "Total Audio Hours",
      "unit": "hours",
      "description": "Total duration of recorded audio in the standard release"
    },
    {
      "name": "Sampling Rate",
      "unit": "kHz",
      "description": "Native or standardized digital audio sampling frequency"
    },
    {
      "name": "Speaker Count",
      "unit": "",
      "description": "Number of unique recorded speakers across all partitions"
    },
    {
      "name": "Acoustic Domain",
      "unit": "",
      "description": "Recording environment, delivery style, and speech type"
    },
    {
      "name": "Primary License",
      "unit": "",
      "description": "Legal redistribution and licensing terms"
    },
    {
      "name": "Key Research Utility",
      "unit": "",
      "description": "Primary benchmark role and modeling application"
    }
  ],
  "rows": [
    {
      "Corpus Name": "LibriSpeech (SLR12)",
      "Total Audio Hours": 1000,
      "Sampling Rate": 16,
      "Speaker Count": 2484,
      "Acoustic Domain": "Read audiobooks (LibriVox)",
      "Primary License": "CC BY 4.0",
      "Key Research Utility": "Supervised ASR benchmarking across clean and other acoustic conditions"
    },
    {
      "Corpus Name": "TIMIT (LDC93S1)",
      "Total Audio Hours": 5.4,
      "Sampling Rate": 16,
      "Speaker Count": 630,
      "Acoustic Domain": "Phonetically balanced read prompts",
      "Primary License": "LDC User Agreement (Proprietary)",
      "Key Research Utility": "Phoneme recognition and acoustic-phonetic modeling"
    },
    {
      "Corpus Name": "Wall Street Journal (WSJ0/WSJ1)",
      "Total Audio Hours": 82,
      "Sampling Rate": 16,
      "Speaker Count": 284,
      "Acoustic Domain": "Read newspaper articles",
      "Primary License": "LDC User Agreement (Proprietary)",
      "Key Research Utility": "Traditional hybrid HMM-GMM acoustic modeling baseline"
    },
    {
      "Corpus Name": "Common Voice (Mozilla)",
      "Total Audio Hours": 30000,
      "Sampling Rate": 48,
      "Speaker Count": 100000,
      "Acoustic Domain": "Crowdsourced read sentences",
      "Primary License": "CC0 1.0 Universal",
      "Key Research Utility": "Multilingual and diverse accent robustness evaluation"
    },
    {
      "Corpus Name": "Libri-Light",
      "Total Audio Hours": 60000,
      "Sampling Rate": 16,
      "Speaker Count": 10000,
      "Acoustic Domain": "Unannotated read audiobooks",
      "Primary License": "CC BY 4.0",
      "Key Research Utility": "Self-supervised representation learning (wav2vec 2.0, HuBERT)"
    },
    {
      "Corpus Name": "Multilingual LibriSpeech (MLS)",
      "Total Audio Hours": 50000,
      "Sampling Rate": 16,
      "Speaker Count": 15000,
      "Acoustic Domain": "Multilingual read audiobooks (8 European languages)",
      "Primary License": "CC BY 4.0",
      "Key Research Utility": "Cross-lingual transfer learning and multilingual speech recognition"
    },
    {
      "Corpus Name": "VoxCeleb1",
      "Total Audio Hours": 350,
      "Sampling Rate": 16,
      "Speaker Count": 1251,
      "Acoustic Domain": "Extracted YouTube interview audio",
      "Primary License": "CC BY 4.0",
      "Key Research Utility": "Speaker identification and speaker verification"
    }
  ],
  "sources": [
    {
      "title": "Panayotov et al., LibriSpeech: An ASR Corpus Based on Public Domain Audio Books (ICASSP 2015)",
      "url": "https://danielpovey.com/files/2015_icassp_librispeech.pdf",
      "accessed": "2026-11-16"
    },
    {
      "title": "OpenSLR Resource SLR12: LibriSpeech ASR Corpus",
      "url": "http://www.openslr.org/12/",
      "accessed": "2026-11-16"
    },
    {
      "title": "Kahn et al., Libri-Light: A Benchmark for ASR with Limited or No Supervision (ICASSP 2020)",
      "url": "https://arxiv.org/abs/1912.07875",
      "accessed": "2026-11-16"
    },
    {
      "title": "Pratap et al., MLS: A Large-Scale Multilingual Dataset for Speech Research (Interspeech 2020)",
      "url": "https://arxiv.org/abs/2012.03411",
      "accessed": "2026-11-16"
    },
    {
      "title": "Garofolo et al., DARPA TIMIT Acoustic-Phonetic Continuous Speech Corpus (LDC93S1)",
      "url": "https://catalog.ldc.upenn.edu/LDC93S1",
      "accessed": "2026-11-16"
    },
    {
      "title": "Ardila et al., Common Voice: A Massively-Multilingual Speech Corpus (LREC 2020)",
      "url": "https://arxiv.org/abs/1912.06670",
      "accessed": "2026-11-16"
    },
    {
      "title": "Nagrani et al., VoxCeleb: A Large-Scale Speaker Identification Dataset (Interspeech 2017)",
      "url": "https://arxiv.org/abs/1706.08612",
      "accessed": "2026-11-16"
    }
  ]
}