Latent Sequence Decompositions
Abstract
We present the Latent Sequence Decompositions (LSD) framework. LSD decomposes sequences with variable lengthed output units as a function of both the input sequence and the output sequence. We present a training algorithm which samples valid extensions and an approximate decoding algorithm. We experiment with the Wall Street Journal speech recognition task. Our LSD model achieves 12.9% WER compared to a character baseline of 14.8% WER. When combined with a convolutional network on the encoder, we achieve 9.6% WER.
Cite
Text
Chan et al. "Latent Sequence Decompositions." International Conference on Learning Representations, 2017.Markdown
[Chan et al. "Latent Sequence Decompositions." International Conference on Learning Representations, 2017.](https://mlanthology.org/iclr/2017/chan2017iclr-latent/)BibTeX
@inproceedings{chan2017iclr-latent,
title = {{Latent Sequence Decompositions}},
author = {Chan, William and Zhang, Yu and Le, Quoc V. and Jaitly, Navdeep},
booktitle = {International Conference on Learning Representations},
year = {2017},
url = {https://mlanthology.org/iclr/2017/chan2017iclr-latent/}
}