English

Understanding Neural Machine Translation by Simplification: The Case of Encoder-free Models

Computation and Language 2019-07-19 v1

Abstract

In this paper, we try to understand neural machine translation (NMT) via simplifying NMT architectures and training encoder-free NMT models. In an encoder-free model, the sums of word embeddings and positional embeddings represent the source. The decoder is a standard Transformer or recurrent neural network that directly attends to embeddings via attention mechanisms. Experimental results show (1) that the attention mechanism in encoder-free models acts as a strong feature extractor, (2) that the word embeddings in encoder-free models are competitive to those in conventional models, (3) that non-contextualized source representations lead to a big performance drop, and (4) that encoder-free models have different effects on alignment quality for German-English and Chinese-English.

Keywords

Cite

@article{arxiv.1907.08158,
  title  = {Understanding Neural Machine Translation by Simplification: The Case of Encoder-free Models},
  author = {Gongbo Tang and Rico Sennrich and Joakim Nivre},
  journal= {arXiv preprint arXiv:1907.08158},
  year   = {2019}
}

Comments

Accepted by RANLP 2019, camera ready version

R2 v1 2026-06-23T10:24:33.576Z