{"success":true,"data":{"author":"Titouan Parcollet, Yuan Tseng, Shucong Zhang, Rogier van Dalen (SpeechBrain, Samsung AI Center Cambridge)","bucketAccessType":"public","bucketName":"speechbrain/LoquaciousSet","bucketPrefix":"","category":"Speech & Audio","cloudProvider":"huggingface","createdAt":"2026-11-14T14:15:18.124Z","description":"LoquaciousSet (also published as LargeScaleASR) is a 25,000-hour corpus of transcribed, heterogeneous English speech for research and commercial speech recognition. It mixes read and spontaneous, clean and noisy speech drawn from five openly licensed corpora: VoxPopuli (CC0), LibriHeavy (CC BY 4.0), LibriSpeech dev-other and test-other (CC BY 4.0), YODAS (CC BY 3.0), People's Speech (CC BY 4.0) and Common Voice 18.0 (CC0), with the mixing and curation described in the accompanying paper. Six subsets are provided: large (25,000 hours), medium (2,500), small (250), clean (13,000 hours, excluding YODAS and People's Speech), dev (15 hours) and test (21 hours). The repository holds those as folders of Parquet shards (the clean folder lists at least 1,000 shard files), plus per-split CSV manifests, a lexicon, a Sequitur G2P model and pruned and unpruned 3-gram and 4-gram ARPA language models (measured 2026-11-14; the Hub reports about 3.8 TB). Load with datasets.load_dataset('speechbrain/LoquaciousSet', 'small') and decode the 'wav' bytes with torchaudio; a SpeechBrain conformer recipe is linked. Every component is open; attribute each source corpus as listed on the card.","displayTitle":"LoquaciousSet (LargeScaleASR): 25,000 hours of English speech","id":"43ea6f4a-fdba-5e2c-beb2-a5017bea940a","license":"CC-BY-4.0","licenseEvidenceUrl":"https://huggingface.co/datasets/speechbrain/LoquaciousSet","objectCountEstimate":9756,"probeStatus":"ok","referenceUrl":"https://arxiv.org/abs/2505.21578","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"speechbrain/LoquaciousSet\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"speechbrain/LoquaciousSet\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=speechbrain, license:cc-by-3.0, size_categories:10M<n<100M, format:parquet, downloads=25684"],"format":"mixed","generatedAt":"2026-11-14T14:15:31.576Z","layout":{"example":"clean/train-00000-of-02615.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"clean/train-00000-of-02615.parquet","truncatedAt":512}],"sampleObjects":["3gram-pruned.arpa.bin","3gram-pruned.arpa.gz","4gram-pruned.arpa.bin","4gram-pruned.arpa.gz","4gram-unpruned.arpa.bin"],"sidecars":[{"bytes":"---\nannotations_creators:\n  - crowdsourced\n  - machine-generated\nlanguage_creators:\n  - crowdsourced\n  - machine-generated\nlanguage:\n  - en\nlicense:\n  - cc-by-3.0\n  - cc-by-4.0\nmultilinguality:\n  - monolingual\nsize_categories:\n  - 1T<n\ntask_categories:\n  - automatic-speech-recognition\ntask_ids: []\npretty_name: LargeScaleASR\ntags:\n  - robust-speech-recognition\n  - noisy-speech-recognition\n  - speech-recognition\nconfigs:\n- config_name: large\n  features:\n  - name: ID\n    dtype: string\n  - name: duration\n    dtype: float32\n  - name: wav\n    dtype:\n      audio:\n        sample_rate: 16000\n        decode: False\n  - name: spk_id\n    dtype: string\n  - name: sex\n    dtype: string\n  - name: text\n    dtype: string\n  data_files:\n  - split: train\n    path: large/train*\n  - split: dev\n    path: dev/dev*\n  - split: test\n    path: test/test*\n- config_name: clean\n  features:\n  - name: ID\n    dtype: string\n  - name: duration\n    dtype: float32\n  - name: wav\n    dtype:\n      audio:\n        sample_rate: 16000\n        decode: False\n  - name: spk_id\n    dtype: string\n  - name: sex\n    dtype: string\n  - name: text\n    dtype: string\n  data_files:\n  - split: train\n    path: clean/train*\n  - split: dev\n    path: dev/dev*\n  - split: test\n    path: test/test*\n- config_name: small\n  features:\n  - name: ID\n    dtype: string\n  - name: duration\n    dtype: float32\n  - name: wav\n    dtype:\n      audio:\n        sample_rate: 16000\n        decode: False\n  - name: spk_id\n    dtype: string\n  - name: sex\n    dtype: string\n  - name: text\n    dtype: string\n  data_files:\n  - split: train\n    path: small/train*\n  - split: dev\n    path: dev/dev*\n  - split: test\n    path: test/test*\n- config_name: medium\n  features:\n  - name: ID\n    dtype: string\n  - name: duration\n    dtype: float32\n  - name: wav\n    dtype:\n      audio:\n        sample_rate: 16000\n        decode: False\n  - name: spk_id\n    dtype: string\n  - name: sex\n    dtype: string\n  - name: text\n    dtype: string\n  data_files:\n  - split: train\n    path: medium/train*\n  - split: dev\n    path: dev/dev*\n  - split: test\n    path: test/test*\n---\n\n# LargeScaleASR: 25,000 hours of transcribed and heterogeneous English speech recognition data for research and commercial use.\n\nThe full details [are available in the paper](https://arxiv.org/abs/2505.21578).\n\nMade of 6 subsets:\n1. **large** contains 25,000 hours of read / spontaneous and clean / noisy transcribed speech.\n2. **medium** contains 2,500 hours of read / spontaneous and clean / noisy transcribed speech.\n3. **small** contains 250 hours of read / spontaneous and clean / noisy transcribed speech.\n4. **clean** contains 13,000 hours of read / spontaneous transcribed speech. YODA and People's Speech data are excluded from this subset as, despite data curation, some errors remain in the transcriptions.\n5. **dev** contains 15 hours (more details in the next section).\n6. **test** contains 21 hours (more details in the next section).\n\nThe large split requires 4TB of storage (including HuggingFace extraction). The shards only are 2TB.\n\n\nExample:\n\n```python\nfrom datasets import load_dataset\n\nds = load_dataset('speechbrain/LoquaciousSet', {'small'||'medium'||'large'}, num_proc={nb_of_cpu_cores_you_want})\nprint(ds['train'])\n\nfrom io import BytesIO\nimport torchaudio\nwav_tensor = torchaudio.load(BytesIO(ds[\"train\"][0][\"wav\"]['bytes']))\n```\n\n## Training recipe\n\nA full conformer ASR training recipe is available [here](https://github.com/speechbrain/speechbrain/pull/2806).\n\n\n## Data description (Following information are directly copy-pasted from the SpeechBrain data preparation README)\n\nTLS is a mix of 5 existing dataset with permissive licences. The way it is mixed\nis described in the following table:\n\n| Dataset       | Amount Taken (large/medium/small/dev/test) | License |\n| ------------- | ------------- | ------------- |\n| VoxPopuli | 550/500/50/5/7 | CC0  |\n| LibriHeavy | 11,000/500/50/0/0 | CC BY 4.0 |\n| Librispeech (dev-/test-other) | 0/0/0/5/7 | CC BY 4.0 |\n| yodas | 6,100/500/50/1.5/1.5 |  CC BY 3.0 |\n| people's speech | 5,900/500/50/1.5/1.5 | CC-BY 4.0 |\n| CommonVoice 18.","encoding":"utf8","key":"README.md","name":"README.md","size":10126,"truncated":true}],"sourceObjectKey":"clean/train-00000-of-02615.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"speechbrain/LoquaciousSet\")\nprint(ds)","starterSamplePath":"3gram-pruned.arpa.bin","totalObjects":9756},"summary":"LoquaciousSet (also published as LargeScaleASR) is a 25,000-hour corpus of transcribed, heterogeneous English speech for research and commercial speech recognition.","title":"LoquaciousSet (LargeScaleASR): 25,000 hours of English speech","updatedAt":"2026-11-14T17:46:18.790Z","visibility":"public","websiteUrl":"https://huggingface.co/datasets/speechbrain/LoquaciousSet"}}