Golos dataset
You can read the Russian version of the documentation.
Golos is a Russian corpus 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. We have made the corpus freely available for downloading, along with the acoustic model prepared on this corpus. Also we create 3-gram KenLM language model using an open Common Crawl corpus.
Dataset structure
| Domain | Train files | Train hours | Test files | Test hours |
|---|---|---|---|---|
| Crowd | 979 796 | 1 095 | 9 994 | 11.2 |
| Farfield | 124 003 | 132.4 | 1 916 | 1.4 |
| Total | 1 103 799 | 1 227.4 | 11 910 | 12.6 |
Downloads
Audio files in opus format
| Archive | Size | Link |
|---|---|---|
| golos_opus.tar | 20.5 Gb | https://sc.link/JpD |
Audio files in wav format
Manifest files with all the training transcription texts are in the train_crowd9.tar archive listed in the table:
| Archives | Size | Links |
|---|---|---|
| train_farfield.tar | 15.4 Gb | https://sc.link/1Z3 |
| train_crowd0.tar | 11 Gb | https://sc.link/Lrg |
| train_crowd1.tar | 14 Gb | https://sc.link/MvQ |
| train_crowd2.tar | 13.2 Gb | https://sc.link/NwL |
| train_crowd3.tar | 11.6 Gb | https://sc.link/Oxg |
| train_crowd4.tar | 15.8 Gb | https://sc.link/Pyz |
| train_crowd5.tar | 13.1 Gb | https://sc.link/Qz7 |
| train_crowd6.tar | 15.7 Gb | https://sc.link/RAL |
| train_crowd7.tar | 12.7 Gb | https://sc.link/VG5 |
| train_crowd8.tar | 12.2 Gb | https://sc.link/WJW |
| train_crowd9.tar | 8.08 Gb | https://sc.link/XKk |
| test.tar | 1.3 Gb | https://sc.link/Kqr |
Acoustic and language models
Acoustic model built using QuartzNet15x5 architecture and trained using NeMo toolkit .
Three n-gram language models created using KenLM Language Model Toolkit .
- LM built on Common Crawl Russian dataset.
- LM built on Golos train set.
- LM built on Common Crawl and Golos datasets together (50/50).
| Archives | Size | Links |
|---|---|---|
| QuartzNet15x5_golos.nemo | 68 MB | https://sc.link/ZMv |
| KenLMs.tar | 4.8 Gb | https://sc.link/YL0 |
Golos data and models are also available in the hub of pre-trained models, datasets, and containers — DataHub ML Space. You can train the model and deploy it on the high-performance SberCloud infrastructure in ML Space — full-cycle machine learning development platform for DS-teams collaboration based on the Christofari Supercomputer.
Evaluation
Percents of Word Error Rate for different test sets.
| Decoder / Test set | Crowd test | Farfield test | MCV dev | MCV test |
|---|---|---|---|---|
| Greedy decoder | 4.389% | 14.949% | 9.314% | 11.278% |
| Beam Search with Common Crawl LM | 4.709% | 12.503% | 6.341% | 7.976% |
| Beam Search with Golos train set LM | 3.548% | 12.384% | — | — |
| Beam Search with Common Crawl and Golos LM | 3.318% | 11.488% | 6.4% | 8.06% |
MCV — Mozilla Common Voice — Mozilla's initiative to help teach machines how real people speak.
Resources
Golos: Russian Dataset for Speech Research .
License
Creators
- Aleksandr Denisenko.
- Angelina Kovalenko.
- Nikolaj Karpov.
- Fedor Min'kin.
Contacts
Ask your questions by email SmartSpeech@sberbank.ru.