Publication Details
Usted: Improving ASR with a Unified Speech and Text Encoder-Decoder
sequence-to-sequence, multitask, end-to-end ASR, masked language model, machine
translation
Improving end-to-end speech recognition by incorporating external text data has
been a longstanding research topic. There has been a recent focus on training E2E
ASR models that get the performance benefits of external text data without
incurring the extra cost of evaluating an external language model at inference
time. In this work, we propose training ASR model jointly with a set of
text-to-text auxiliary tasks with which it shares a decoder and parts of the
encoder. When we jointly train ASR and masked language model with the 960-hour
Librispeech and Opensubtitles data respectively, we observe WER reductions of 16%
and 20% on test-other and test-clean respectively over an ASR-only baseline
without any extra cost at inference time, and reductions of 6% and 8% compared to
a stronger MUTE-L baseline which trains the decoder with the same text data as
our model. We achieve further improvements when we train masked language model on
Librispeech data or when we use machine translation as the auxiliary task,
without significantly sacrificing performance on the task itself.
@inproceedings{BUT178379,
author="Bolaji {Yusuf} and Ankur {Gandhe} and Alex {Sokolov}",
title="Usted: Improving ASR with a Unified Speech and Text Encoder-Decoder",
booktitle="ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",
year="2022",
pages="8297--8301",
publisher="IEEE Signal Processing Society",
address="Singapore",
doi="10.1109/ICASSP43922.2022.9746554",
isbn="978-1-6654-0540-9",
url="https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9746554"
}