Data and Parameter Scaling Laws for Neural Machine Translation

Mitchell A Gordon, Kevin Duh, Jared Kaplan


Abstract
We observe that the development cross-entropy loss of supervised neural machine translation models scales like a power law with the amount of training data and the number of non-embedding parameters in the model. We discuss some practical implications of these results, such as predicting BLEU achieved by large scale models and predicting the ROI of labeling data in low-resource language pairs.
Anthology ID:
2021.emnlp-main.478
Volume:
Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing
Month:
November
Year:
2021
Address:
Online and Punta Cana, Dominican Republic
Editors:
Marie-Francine Moens, Xuanjing Huang, Lucia Specia, Scott Wen-tau Yih
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
5915–5922
Language:
URL:
https://preview.aclanthology.org/build-pipeline-with-new-library/2021.emnlp-main.478/
DOI:
10.18653/v1/2021.emnlp-main.478
Bibkey:
Cite (ACL):
Mitchell A Gordon, Kevin Duh, and Jared Kaplan. 2021. Data and Parameter Scaling Laws for Neural Machine Translation. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 5915–5922, Online and Punta Cana, Dominican Republic. Association for Computational Linguistics.
Cite (Informal):
Data and Parameter Scaling Laws for Neural Machine Translation (Gordon et al., EMNLP 2021)
Copy Citation:
PDF:
https://preview.aclanthology.org/build-pipeline-with-new-library/2021.emnlp-main.478.pdf
Video:
 https://preview.aclanthology.org/build-pipeline-with-new-library/2021.emnlp-main.478.mp4
Data
OpenSubtitles