An Empirical Study of Mini-Batch Creation Strategies for Neural Machine Translation

Makoto Morishita, Yusuke Oda, Graham Neubig, Koichiro Yoshino, Katsuhito Sudoh, Satoshi Nakamura


Abstract
Training of neural machine translation (NMT) models usually uses mini-batches for efficiency purposes. During the mini-batched training process, it is necessary to pad shorter sentences in a mini-batch to be equal in length to the longest sentence therein for efficient computation. Previous work has noted that sorting the corpus based on the sentence length before making mini-batches reduces the amount of padding and increases the processing speed. However, despite the fact that mini-batch creation is an essential step in NMT training, widely used NMT toolkits implement disparate strategies for doing so, which have not been empirically validated or compared. This work investigates mini-batch creation strategies with experiments over two different datasets. Our results suggest that the choice of a mini-batch creation strategy has a large effect on NMT training and some length-based sorting strategies do not always work well compared with simple shuffling.
Anthology ID:
W17-3208
Volume:
Proceedings of the First Workshop on Neural Machine Translation
Month:
August
Year:
2017
Address:
Vancouver
Editors:
Thang Luong, Alexandra Birch, Graham Neubig, Andrew Finch
Venue:
NGT
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
61–68
Language:
URL:
https://aclanthology.org/W17-3208
DOI:
10.18653/v1/W17-3208
Bibkey:
Cite (ACL):
Makoto Morishita, Yusuke Oda, Graham Neubig, Koichiro Yoshino, Katsuhito Sudoh, and Satoshi Nakamura. 2017. An Empirical Study of Mini-Batch Creation Strategies for Neural Machine Translation. In Proceedings of the First Workshop on Neural Machine Translation, pages 61–68, Vancouver. Association for Computational Linguistics.
Cite (Informal):
An Empirical Study of Mini-Batch Creation Strategies for Neural Machine Translation (Morishita et al., NGT 2017)
Copy Citation:
PDF:
https://aclanthology.org/W17-3208.pdf
Data
WMT 2016