This model transcribes speech into lowercase Cyrillic alphabet including space, and is trained on around 1636 hours of Russian speech data.
It is a non-autoregressive "large" variant of Conformer, with around 120 million parameters.
See the model architecture section and NeMo documentation for complete architecture details.
Usage
The model is available for use in the NeMo toolkit [3], and can be used as a pre-trained checkpoint for inference or for fine-tuning on another dataset.
To train, fine-tune or play with the model you will need to install NVIDIA NeMo. We recommend you install it after you've installed latest PyTorch version.
pip install nemo_toolkit['all']
Automatically instantiate the model
python
1import nemo.collections.asr as nemo_asr
2asr_model = nemo_asr.models.EncDecRNNTBPEModel.from_pretrained("nvidia/stt_ru_conformer_transducer_large")
This model accepts 16 kHz mono-channel Audio (wav files) as input.
Output
This model provides transcribed speech as a string for a given audio sample.
Model Architecture
Conformer-Transducer model is an autoregressive variant of Conformer model [1] for Automatic Speech Recognition which uses Transducer loss/decoding. You may find more info on the detail of this model here: Conformer-Transducer Model.
Training
The NeMo toolkit [3] was used for training the models for over several hundred epochs. These model are trained with this example script and this base config.
Rare symbols with diacritics were replaced during preprocessing.
The tokenizers for these models were built using the text transcripts of the train set with this script.
Datasets
All the models in this collection are trained on a composite dataset (NeMo ASRSET) comprising of more than a thousand hours of Russian speech:
Mozilla Common Voice 10.0 (Russian) - train subset [28 hours]
Golos - crowd [1070 hours] and fairfield [111 hours] subsets
Russian LibriSpeech (RuLS) [92 hours]
SOVA - RuAudiobooksDevices [260 hours] and RuDevices [75 hours] subsets
Performance
The list of the available models in this collection is shown in the following table. Performances of the ASR models are reported in terms of Word Error Rate (WER%) with greedy decoding.
Version
Tokenizer
Vocabulary Size
MCV 10.0 dev
MCV 10.0 test
GOLOS-crowd test
GOLOS-farfield test
RuLS test
Train Dataset
1.13.0
SentencePiece Unigram
1024
3.5
4.0
2.7
7.6
12.0
NeMo ASRSET
Limitations
Since this model was trained on publicly available speech datasets, the performance of this model might degrade for speech which includes technical terms, or vernacular that the model has not been trained on. The model might also perform worse for accented speech.
Deployment with NVIDIA Riva
NVIDIA Riva, is an accelerated speech AI SDK deployable on-prem, in all clouds, multi-cloud, hybrid, on edge, and embedded.
Additionally, Riva provides:
World-class out-of-the-box accuracy for the most common languages with model checkpoints trained on proprietary data with hundreds of thousands of GPU-compute hours
Best in class accuracy with run-time word boosting (e.g., brand and product names) and customization of acoustic model, language model, and inverse text normalization
Streaming speech recognition, Kubernetes compatible scaling, and enterprise-grade support
