Word-Level Speech Recognition With a Letter to Word Encoder
Ronan Collobert, Awni Hannun, Gabriel Synnaeve
Abstract
We propose a direct-to-word sequence model which uses a word network to learn word embeddings from letters. The word network can be integrated seamlessly with arbitrary sequence models including Connectionist Temporal Classification and encoder-decoder models with attention. We show our direct-to-word model can achieve word error rate gains over sub-word level models for speech recognition. We also show that our direct-to-word approach retains the ability to predict words not seen at training time without any retraining. Finally, we demonstrate that a word-level model can use a larger stride than a sub-word level model while maintaining accuracy. This makes the model more efficient both for training and inference.
BibTeX
@InProceedings{pmlr-v119-collobert20a,
title = {Word-Level Speech Recognition With a Letter to Word Encoder},
author = {Collobert, Ronan and Hannun, Awni and Synnaeve, Gabriel},
booktitle = {Proceedings of the 37th International Conference on Machine Learning},
pages = {2100--2110},
year = {2020},
editor = {III, Hal Daumé and Singh, Aarti},
volume = {119},
series = {Proceedings of Machine Learning Research},
month = {13--18 Jul},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v119/collobert20a/collobert20a.pdf},
url = {https://proceedings.mlr.press/v119/collobert20a.html},
abstract = {We propose a direct-to-word sequence model which uses a word network to learn word embeddings from letters. The word network can be integrated seamlessly with arbitrary sequence models including Connectionist Temporal Classification and encoder-decoder models with attention. We show our direct-to-word model can achieve word error rate gains over sub-word level models for speech recognition. We also show that our direct-to-word approach retains the ability to predict words not seen at training time without any retraining. Finally, we demonstrate that a word-level model can use a larger stride than a sub-word level model while maintaining accuracy. This makes the model more efficient both for training and inference.}
}