Understanding Neural Machine Translation by Simplification: The Case of Encoder-free Models
arXiv:1907.08158
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.
arXiv commentAccepted by RANLP 2019, camera ready version
References in corpus (5)
- Sequence to Sequence Learning with Neural Networks
- MXNet: A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems
- What do Neural Machine Translation Models Learn about Morphology?
- Sockeye: A Toolkit for Neural Machine Translation
- Adding Interpretable Attention to Neural Translation Models Improves Word Alignment