Learning Deep Transformer Models for Machine Translation

TLDR

Transformer is the state‑of‑the‑art model for machine translation, and research to improve it focuses on either widening networks (Transformer‑Big) or deepening language representations, the latter facing challenges in training deep models. The study aims to advance deep Transformer models. The authors propose a deep Transformer that uses layer normalization and a novel residual combination of previous layers to improve performance. The deep model achieved 0.4‑2.4 BLEU point gains over Transformer‑Big on WMT’16 English‑German and NIST OpenMT’12 Chinese‑English, while being 1.6× smaller and 3× faster to train.

Abstract

Transformer is the state-of-the-art model in recent machine translation evaluations. Two strands of research are promising to improve models of this kind: the first uses wide networks (a.k.a. Transformer-Big) and has been the de facto standard for development of the Transformer system, and the other uses deeper language representation but faces the difficulty arising from learning deep networks. Here, we continue the line of research on the latter. We claim that a truly deep Transformer model can surpass the Transformer-Big counterpart by 1) proper use of layer normalization and 2) a novel way of passing the combination of previous layers to the next. On WMT’16 English-German and NIST OpenMT’12 Chinese-English tasks, our deep system (30/25-layer encoder) outperforms the shallow Transformer-Big/Base baseline (6-layer encoder) by 0.4-2.4 BLEU points. As another bonus, the deep model is 1.6X smaller in size and 3X faster in training than Transformer-Big.