This repository provides all the necessary tools to perform automatic speech recognition from an end-to-end system pretrained on CommonVoice (Spanish Language) within SpeechBrain. For a better experience, we encourage you to learn more about SpeechBrain.

The performance of the model is the following:

Release	RTVE 2022 Test WER	GPUs
16-01-23	23.45	3xRTX2080Ti 12GB

Pipeline description

This ASR system is composed of 2 different but linked blocks:

Tokenizer (char) that transforms words into chars and trained with the train transcriptions (train.tsv) of CommonVoice (ES).
Acoustic model (wav2vec2.0 + CTC). A pretrained wav2vec 2.0 model (wav2vec2-large-xlsr-53-spanish) is combined with two DNN layers and finetuned on CommonVoice ES. The obtained final acoustic representation is given to the CTC decoder.

The system is trained with recordings sampled at 16kHz (single channel). The code will automatically normalize your audio (i.e., resampling + mono channel selection) when calling transcribe_file if needed.

Install SpeechBrain

First of all, please install tranformers and SpeechBrain with the following command:

pip install speechbrain transformers

Please notice that we encourage you to read tutorials and learn more about SpeechBrain.

Transcribing your own audio files (in Spanish)

python

from speechbrain.pretrained import EncoderASR
asr_model = EncoderASR.from_hparams(source="Voyager1/asr-wav2vec2-commonvoice-es", savedir="pretrained_models/asr-wav2vec2-commonvoice-es")
asr_model.transcribe_file("Voyager1/asr-wav2vec2-commonvoice-es/example-es.wav")

Inference on GPU

To perform inference on the GPU, add run_opts={"device":"cuda"} when calling the from_hparams method.

Limitations

We do not provide any warranty on the performance achieved by this model when used on other datasets.

Citations

bibtex

@article{lopez2022tid,
  title={TID Spanish ASR system for the Albayzin 2022 Speech-to-Text Transcription Challenge},
  author={L{\'o}pez, Fernando and Luque, Jordi},
  journal={Proc. IberSPEECH 2022},
  pages={271--275},
  year={2022}
}
@misc{https://doi.org/10.48550/arxiv.2210.15226,
  doi = {10.48550/ARXIV.2210.15226},
  url = {https://arxiv.org/abs/2210.15226},
  author = {López, Fernando and Luque, Jordi},
  title = {Iterative pseudo-forced alignment by acoustic CTC loss for self-supervised ASR domain adaptation},
  publisher = {arXiv},
  year = {2022},
  copyright = {Creative Commons Attribution 4.0 International}
}
@misc{lleidartve,
  title={Rtve 2018, 2020 and 2022 database description},
  author={Lleida, E and Ortega, A and Miguel, A and Baz{\'a}n, V and P{\'e}rez, C and G{\'o}mez, M and de Prada, A}
}
@misc{speechbrain,
  title={{SpeechBrain}: A General-Purpose Speech Toolkit},
  author={Mirco Ravanelli and Titouan Parcollet and Peter Plantinga and Aku Rouhe and Samuele Cornell and Loren Lugosch and Cem Subakan and Nauman Dawalatabad and Abdelwahab Heba and Jianyuan Zhong and Ju-Chieh Chou and Sung-Lin Yeh and Szu-Wei Fu and Chien-Feng Liao and Elena Rastorgueva and François Grondin and William Aris and Hwidong Na and Yan Gao and Renato De Mori and Yoshua Bengio},
  year={2021},
  eprint={2106.04624},
  archivePrefix={arXiv},
  primaryClass={eess.AS},
  note={arXiv:2106.04624}
}

Voyager1/asr-wav2vec2-commonvoice-es-finetuned-rtve

wav2vec 2.0 with CTC trained on data aligned from RTVE databases (No LM)