English

BERT, mBERT, or BiBERT? A Study on Contextualized Embeddings for Neural Machine Translation

Computation and Language 2021-09-13 v1

Abstract

The success of bidirectional encoders using masked language models, such as BERT, on numerous natural language processing tasks has prompted researchers to attempt to incorporate these pre-trained models into neural machine translation (NMT) systems. However, proposed methods for incorporating pre-trained models are non-trivial and mainly focus on BERT, which lacks a comparison of the impact that other pre-trained models may have on translation performance. In this paper, we demonstrate that simply using the output (contextualized embeddings) of a tailored and suitable bilingual pre-trained language model (dubbed BiBERT) as the input of the NMT encoder achieves state-of-the-art translation performance. Moreover, we also propose a stochastic layer selection approach and a concept of dual-directional translation model to ensure the sufficient utilization of contextualized embeddings. In the case of without using back translation, our best models achieve BLEU scores of 30.45 for En->De and 38.61 for De->En on the IWSLT'14 dataset, and 31.26 for En->De and 34.94 for De->En on the WMT'14 dataset, which exceeds all published numbers.

Keywords

Cite

@article{arxiv.2109.04588,
  title  = {BERT, mBERT, or BiBERT? A Study on Contextualized Embeddings for Neural Machine Translation},
  author = {Haoran Xu and Benjamin Van Durme and Kenton Murray},
  journal= {arXiv preprint arXiv:2109.04588},
  year   = {2021}
}

Comments

EMNLP 2021

R2 v1 2026-06-24T05:50:40.591Z