{"url":"/dataset/golos","name":"GOLOS","full_name":null,"description_markdown":"**Golos** is a Russian speech dataset suitable for speech research. The dataset mainly consists of recorded audio files manually annotated on the crowd-sourcing platform. The total duration of the audio is about 1240 hours.\r\n\r\n\r\n## **Dataset structure**\r\n\r\n| Domain         | Train files | Train hours  | Test files | Test hours |\r\n|----------------|------------|--------|-------|------|\r\n| Crowd          | 979 796    | 1 095  | 9 994 | 11.2 |\r\n| Farfield       | 124 003    |   132.4| 1 916 |  1.4 |\r\n| Total          | 1 103 799  | 1 227.4|11 910 | 12.6 |\r\n\r\n\r\n\r\n\r\n### **Audio files in opus format**\r\n\r\n| Archive          | Size       |  Link               |\r\n|------------------|------------|---------------------|\r\n| golos_opus.tar   | 20.5 GB    | https://sc.link/JpD |\r\n\r\n\r\n\r\n### **Audio files in wav format**\r\n\r\n| Archives          | Size       |  Links              |\r\n|-------------------|------------|---------------------|\r\n| train_farfield.tar| 15.4 GB    | https://sc.link/1Z3 |\r\n| train_crowd0.tar  | 11 GB      | https://sc.link/Lrg |\r\n| train_crowd1.tar  | 14 GB      | https://sc.link/MvQ |\r\n| train_crowd2.tar  | 13.2 GB    | https://sc.link/NwL |\r\n| train_crowd3.tar  | 11.6 GB    | https://sc.link/Oxg |\r\n| train_crowd4.tar  | 15.8 GB    | https://sc.link/Pyz |\r\n| train_crowd5.tar  | 13.1 GB    | https://sc.link/Qz7 |\r\n| train_crowd6.tar  | 15.7 GB    | https://sc.link/RAL |\r\n| train_crowd7.tar  | 12.7 GB    | https://sc.link/VG5 |\r\n| train_crowd8.tar  | 12.2 GB    | https://sc.link/WJW |\r\n| train_crowd9.tar  | 8.08 GB    | https://sc.link/XKk |\r\n| test.tar          | 1.3 GB     | https://sc.link/Kqr |\r\n\r\n\r\n\r\n## **Evaluation**\r\n\r\nPercents of Word Error Rate for different test sets\r\n\r\n\r\n| Decoder \\ Test set    | Crowd test  | Farfield test    | MCV<sup>1</sup> dev | MCV<sup>1</sup> test |\r\n|-------------------------------------|-----------|----------|-----------|----------|\r\n| Greedy decoder                      | 4.389 %   | 14.949 % | 9.314 %   | 11.278 % |\r\n| Beam Search with Common Crawl LM    | 4.709 %   | 12.503 % | 6.341 %   | 7.976 % |\r\n| Beam Search with Golos train set LM | 3.548 %   | 12.384 % |  -        | -       |\r\n| Beam Search with Common Crawl and Golos LM | 3.318 %   | 11.488 % | 6.4 %     | 8.06 %   |","description_withheld":null,"homepage":"https://github.com/sberdevices/golos","introduced_date":"2021-06-18","introduced_date_note":null,"introduced_by":{"paper":"/paper/golos-russian-dataset-for-speech-research","title":"Golos: Russian Dataset for Speech Research","first_author":"Nikolay Karpov","url":null},"license":{"name":"Custom","url":"https://github.com/sberdevices/golos/blob/master/license/en_us.pdf"},"modalities":[{"name":"Speech","url":"/datasets/modality/speech"}],"tasks":[{"name":"Speech Recognition","url":"/task/speech-recognition","datasets_with_task":"/datasets/task/speech-recognition"}],"languages":[{"name":"Russian","url":"/datasets/language/russian"}],"variants":["GOLOS"],"data_loaders":[{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/bond005/sberdevices_golos_100h_farfield","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/bond005/sberdevices_golos_10h_crowd","frameworks":["tf","pytorch","jax"]}],"num_papers_in_archive":2,"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28"},"benchmarks":[],"papers_with_a_benchmark_row":[],"syntology_totals":{"read_at":"2026-09-24T18:15:14+00:00","papers_with_samples":0,"samples_harvested":0,"samples_ran":0,"samples_unverified":0,"pointer_only_for_licence":0,"papers_with_no_sample_that_ran":0,"note":"the per-paper counts above, summed; not a rate"},"papers_note":"The archive never published its papers-using-dataset list; these are papers with a leaderboard row on this dataset's benchmarks."}