SRPOL’s System for the IWSLT 2020 End-to-End Speech Translation Task

Tomasz Potapczyk, Pawel Przybysz


Abstract
We took part in the offline End-to-End English to German TED lectures translation task. We based our solution on our last year’s submission. We used a slightly altered Transformer architecture with ResNet-like convolutional layer preparing the audio input to Transformer encoder. To improve the model’s quality of translation we introduced two regularization techniques and trained on machine translated Librispeech corpus in addition to iwslt-corpus, TEDLIUM2 andMust_C corpora. Our best model scored almost 3 BLEU higher than last year’s model. To segment 2020 test set we used exactly the same procedure as last year.
Anthology ID:
2020.iwslt-1.9
Volume:
Proceedings of the 17th International Conference on Spoken Language Translation
Month:
July
Year:
2020
Address:
Online
Venues:
ACL | IWSLT
SIG:
SIGSLT
Publisher:
Association for Computational Linguistics
Note:
Pages:
89–94
Language:
URL:
https://aclanthology.org/2020.iwslt-1.9
DOI:
10.18653/v1/2020.iwslt-1.9
Bibkey:
Cite (ACL):
Tomasz Potapczyk and Pawel Przybysz. 2020. SRPOL’s System for the IWSLT 2020 End-to-End Speech Translation Task. In Proceedings of the 17th International Conference on Spoken Language Translation, pages 89–94, Online. Association for Computational Linguistics.
Cite (Informal):
SRPOL’s System for the IWSLT 2020 End-to-End Speech Translation Task (Potapczyk & Przybysz, IWSLT 2020)
Copy Citation:
PDF:
https://aclanthology.org/2020.iwslt-1.9.pdf