Model Overview
This collection contains large size versions of Conformer-CTC (around 120M parameters) trained on LibriSpeech dataset. The model transcribes speech in lower case english alphabet along with spaces and apostrophes.
Trained or fine-tuned NeMo models (with the file extenstion .nemo) can be converted to Riva models (with the file extension .riva) and then deployed. Here is a pre-trained Conformer-CTC speech-to-text (STT) -- a.k.a. automatic speech recognition (ASR) -- Riva model.
Model Architecture
Conformer-CTC model is a non-autoregressive variant of Conformer model [1] for Automatic Speech Recognition which uses CTC loss/decoding instead of Transducer. You may find more info on the detail of this model here: Conformer-CTC 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.
The tokenizers for these models were built using the text transcripts of the train set with this script.
The checkpoint of the language model used as the neural rescorer can be found here. You may find more info on how to train and use language models for ASR models here: ASR Language Modeling
Datasets
All the models in this collection are trained on LibriSpeech dataset which contains around 1000 hours of English speech.
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 | LS test-other | LS test-clean |
|---|---|---|---|---|
| 1.1.0 | SentencePiece Unigram [2] | 128 | 5.7 | 2.5 |
| 1.0.0 | SentencePiece Unigram [2] | 128 | 6.1 | 2.7 |
You may use language models to improve the accuracy of the models. The WER(%) of the latest model with different language modeling techniques are reported in the follwoing table.
| Language Modeling | Training Dataset | LS test-other | LS test-clean | Comment |
|---|---|---|---|---|
| N-gram LM | LS Train + LS LM Corpus | 4.9 | 2.2 | N=10, beam_width=128, n_gram_alpha=2.0, n_gram_beta=1.0 |
| Neural Rescorer(Transformer) | LS Train + LS LM Corpus | 4.8 | 2.0 | N=10, beam_width=128 |
| N-gram + Neural Rescorer(Transformer) | LS Train + LS LM Corpus | 4.5 | 2.0 | N=10, beam_width=128, n_gram_alpha=2.0, n_gram_beta=1.0 |
How to Use this Model
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.
Automatically load the model from NGC
import nemo.collections.asr as nemo_asr
asr_model = nemo_asr.models.EncDecCTCModelBPE.from_pretrained(model_name="stt_en_conformer_ctc_large_ls")
Transcribing text with this model
python [NEMO_GIT_FOLDER]/examples/asr/transcribe_speech.py \
pretrained_name="stt_en_conformer_ctc_large_ls" \
audio_dir=""
Input
This model accepts 16000 KHz Mono-channel Audio (wav files) as input.
Output
This model provides transcribed speech as a string for a given audio sample.
Limitations
Since all models are trained on just LibriSpeech dataset, 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.
References
[1] Conformer: Convolution-augmented Transformer for Speech Recognition
[2] Google Sentencepiece Tokenizer
Licence
License to use this model is covered by the NGC TERMS OF USE unless another License/Terms Of Use/EULA is clearly specified. By downloading the public and release version of the model, you accept the terms and conditions of the NGC TERMS OF USE.