Publication Details

Transfer Learning Of Language-independent End-to-end ASR With Language Model Fusion

INAGUMA Hirofumi, CHO Jaejin, BASKAR Murali K., KAWAHARA Tatsuya and WATANABE Shinji. Transfer Learning Of Language-independent End-to-end ASR With Language Model Fusion. In: Proceedings of ICASSP. Brighton: IEEE Signal Processing Society, 2019, pp. 6096-6100. ISBN 978-1-5386-4658-8. Available from: https://ieeexplore.ieee.org/document/8682918
Czech title
Přenosové učení na jazyce nezávislého end-to-end ASR s fúzí jazykových modelů
Type
conference paper
Language
english
Authors
Inaguma Hirofumi (KyotoUni)
Cho Jaejin (JHU)
Baskar Murali K. (DCGM FIT BUT)
Kawahara Tatsuya (KyotoUni)
Watanabe Shinji, Dr. (JHU)
URL
Keywords

end-to-end ASR, multilingual speech recognition, low-resource language, transfer learning

Abstract

This work explores better adaptation methods to low-resource lan-guages using an external language model (LM) under the frame-work of transfer learning. We first build a language-independentASR system in a unified sequence-to-sequence (S2S) architecturewith a shared vocabulary among all languages. During adaptation,we performLM fusion transfer, where an external LM is integratedinto the decoder network of the attention-based S2S model in thewhole adaptation stage, to effectively incorporate linguistic contextof the target language. We also investigate various seed models fortransfer learning. Experimental evaluations using the IARPA BA-BEL data set show that LM fusion transfer improves performanceson all target five languages compared with simple transfer learningwhen the external text data is available. Our final system drasticallyreduces the performance gap from the hybrid systems.

Published
2019
Pages
6096-6100
Proceedings
Proceedings of ICASSP
Conference
2019 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), Brighton, GB
ISBN
978-1-5386-4658-8
Publisher
IEEE Signal Processing Society
Place
Brighton, GB
DOI
UT WoS
000482554006065
EID Scopus
BibTeX
@INPROCEEDINGS{FITPUB12095,
   author = "Hirofumi Inaguma and Jaejin Cho and K. Murali Baskar and Tatsuya Kawahara and Shinji Watanabe",
   title = "Transfer Learning Of Language-independent End-to-end ASR With Language Model Fusion",
   pages = "6096--6100",
   booktitle = "Proceedings of ICASSP",
   year = 2019,
   location = "Brighton, GB",
   publisher = "IEEE Signal Processing Society",
   ISBN = "978-1-5386-4658-8",
   doi = "10.1109/ICASSP.2019.8682918",
   language = "english",
   url = "https://www.fit.vut.cz/research/publication/12095"
}
Back to top