End-to-End Offline Speech Translation System for IWSLT 2020 using Modality Agnostic Meta-Learning

Nikhil Kumar Lakumarapu, Beomseok Lee, Sathish Reddy Indurthi, Hou Jeung Han, Mohd Abbas Zaidi, Sangha Kim


Abstract
In this paper, we describe the system submitted to the IWSLT 2020 Offline Speech Translation Task. We adopt the Transformer architecture coupled with the meta-learning approach to build our end-to-end Speech-to-Text Translation (ST) system. Our meta-learning approach tackles the data scarcity of the ST task by leveraging the data available from Automatic Speech Recognition (ASR) and Machine Translation (MT) tasks. The meta-learning approach combined with synthetic data augmentation techniques improves the model performance significantly and achieves BLEU scores of 24.58, 27.51, and 27.61 on IWSLT test 2015, MuST-C test, and Europarl-ST test sets respectively.
Anthology ID:
2020.iwslt-1.7
Volume:
Proceedings of the 17th International Conference on Spoken Language Translation
Month:
July
Year:
2020
Address:
Online
Venues:
ACL | IWSLT | WS
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
73–79
Language:
URL:
https://www.aclweb.org/anthology/2020.iwslt-1.7
DOI:
10.18653/v1/2020.iwslt-1.7
Bib Export formats:
BibTeX MODS XML EndNote
PDF:
http://aclanthology.lst.uni-saarland.de/2020.iwslt-1.7.pdf
Video:
 http://slideslive.com/38929596