We explore strategies for incorporating target syntax into Neural Machine Translation. We specifically focus on syntax in ensembles containing multiple sentence representations. We formulate beam search over such ensembles using WFSTs, and describe a delayed SGD update training procedure that is especially effective for long representations like linearized syntax. Our approach gives state-of-the-art performance on a difficult Japanese-English task.
@article{arxiv.1805.00456,
title = {Multi-representation Ensembles and Delayed SGD Updates Improve Syntax-based NMT},
author = {Danielle Saunders and Felix Stahlberg and Adria de Gispert and Bill Byrne},
journal= {arXiv preprint arXiv:1805.00456},
year = {2018}
}