NGC | Catalog
CatalogModelsSTT Pl FastConformer Hybrid Transducer-CTC Large P&C

STT Pl FastConformer Hybrid Transducer-CTC Large P&C

For downloads and more information, please view on a desktop device.
Logo for STT Pl FastConformer Hybrid Transducer-CTC Large P&C

Description

This collection contains the large version (114M) of the Polish speech recognition model with a FastConformer encoder and a Hybrid decoder (joint RNNT-CTC loss). The model has a vocab size of 1024 and emits text with punctuation and capitalization.

Publisher

NVIDIA

Latest Version

1.18.0

Modified

May 2, 2023

Size

437.98 MB

Model Overview

This collection contains the Polish FastConformer Hybrid (Transducer and CTC) Large model (around 114M parameters) with Punctuation and Capitalization. It is trained on the NeMo PnC PL ASR set (MCV, MLS, Voxpopuli) that contains ~165 hours of Polish speech.

It utilizes a Google SentencePiece Unigram [1] tokenizer with vocabulary size 1024, and transcribes text in upper and lower case English alphabet along with spaces, periods, commas, question marks and a few other characters.

Model Architecture

FastConformer is an optimized version of the Conformer model [2] with 8x depthwise-separable convolutional downsampling. The model is trained in a multitask setup with joint Transducer and CTC decoder loss. You may find more information on the details of FastConformer here: Fast-Conformer Model and about Hybrid Transducer-CTC training here: Hybrid Transducer-CTC.

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.

Datasets

All the models in this collection are trained on a composite dataset (NeMo PnC ASRSET) comprising of 165 hours of Polish speech:

  • Mozilla Common Voice 12 (MCV; 45.6 hrs)
  • Multilingual LibriSpeech (MLS; 44.1 hrs)
  • VoxPopuli (75.3 hrs)

Tokenizer Construction

The tokenizer for this model was built using text corpus provided with the train dataset.

We build a Google SentencePiece Tokenizer [1] with the following script :

python [NEMO_GIT_FOLDER]/scripts/tokenizers/process_asr_text_tokenizer.py \
  --manifest="train_manifest.json" \
  --data_root="<OUTPUT DIRECTORY FOR TOKENIZER>" \
  --vocab_size=1024 \
  --tokenizer="spe" \
  --spe_type="unigram" \
  --spe_character_coverage=1.0 \
  --no_lower_case \
  --log

Performance

The performance of Automatic Speech Recognition models is measuring using Word Error Rate. Since this dataset is trained on multiple domains and a much larger corpus, it will generally perform better at transcribing audio in general.

The following tables summarizes the performance of the available models in this collection with the Transducer decoder. Performances of the ASR models are reported in terms of Word Error Rate (WER%) with greedy decoding.

a) On data without Punctuation and Capitalization with Transducer decoder

Version Tokenizer Vocabulary Size MCV12 DEV MCV12 TEST MLS DEV MLS TEST VOXPOPULI DEV VOXPOPULI TEST
1.18.0 SentencePiece Unigram 1024 5.99 8.67 7.15 5.78 11.29 8.45

b) On data with Punctuation and Capitalization with Transducer decoder

Version Tokenizer Vocabulary Size MCV12 DEV MCV12 TEST MLS DEV MLS TEST VOXPOPULI DEV VOXPOPULI TEST
1.18.0 SentencePiece Unigram 1024 8.92 10.98 15.98 11.02 14.03 11.42

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.EncDecHybridRNNTCTCBPEModel.from_pretrained(model_name="stt_pl_fastconformer_hybrid_large_pc")

Transcribing text with this model

Using Transducer mode inference:

python [NEMO_GIT_FOLDER]/examples/asr/transcribe_speech.py \
  pretrained_name="stt_de_fastconformer_hybrid_large_pc" \
  audio_dir="<DIRECTORY CONTAINING AUDIO FILES>"

Using CTC mode inference:

python [NEMO_GIT_FOLDER]/examples/asr/transcribe_speech.py \
  pretrained_name="stt_de_fastconformer_hybrid_large_pc" \
  audio_dir="<DIRECTORY CONTAINING AUDIO FILES>" \
  decoder_type="ctc"

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 this model was trained on publically 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. The model only outputs the punctuations: '.', ',', '?' and hence might not do well in scenarios where other punctuations are also expected.

References

[1] Google Sentencepiece Tokenizer

[2] Conformer: Convolution-augmented Transformer for Speech Recognition

[3] NVIDIA NeMo Toolkit

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.