Pretraining by Backtranslation for End-to-End ASR in Low-Resource Settings

By Matthew Wiesner, Adithya Renduchintala, Shinji Watanabe, Chunxi Liu, Najim Dehak, Sanjeev Khudanpur. 2018

7 citations recorded by the source corpus.

We explore training attention-based encoder-decoder ASR in low-resource settings. These models perform poorly when trained on small amounts of transcribed speech, in part because they depend on having sufficient target-side text to train the attention and decoder networks. In this paper we address this shortcoming by pretraining our network parameters using only text-based data and transcribed speech from other languages. We analyze the relative contributions of both sources of data. Across 3 test languages, our text-based approach resulted in a 20% average relative improvement over a text-based augmentation technique without pretraining. Using transcribed speech from nearby languages gives a further 20-30% relative reduction in character error rate.

Subfields: Machine Translation, Speech & Audio

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Machine Translation

1. Juan Pino
2. Changhan Wang
3. Jiatao Gu
4. Anne Wu
5. Yun Tang
6. Ye Jia
7. Xutai Ma
8. Ann Lee

In Speech & Audio

1. Changhan Wang
2. Juan Pino
3. Jiatao Gu
4. Andrei Andrusenko
5. Aleksandr Laptev
6. Ivan Medennikov
7. Alexis Conneau
8. Ye Jia

Citations in the corpus

• You Do Not Need More Data: Improving End-To-End Speech Recognition by Text-To-Speech Data Augmentation (2020, background): “However, data augmentation techniques can help to make the end-to-end approach more competitive in the low-resource case [4].”
• Improving Cross-Lingual Transfer Learning for End-to-End Speech Recognition with Speech Translation (2020, methodology): “Unlike previous ideas for leveraging translation data [20, 21, 22], our approach does not require any modification to the ASR model architecture.”