In this paper, we present Neural Phrase-based Machine Translation (NPMT). Our method explicitly models the phrase structures in output sequences using Sleep-WAke Networks (SWAN), a recently proposed segmentation-based sequence modeling method. To mitigate the monotonic alignment requirement of SWAN, we introduce a new layer to perform (soft) local reordering of input sequences. Different from existing neural machine translation (NMT) approaches, NPMT does not use attention-based decoding mechanisms. Instead, it directly outputs phrases in a sequential order and can decode in linear time. Our experiments show that NPMT achieves superior performances on IWSLT 2014 German-English/English-German and IWSLT 2015 English-Vietnamese machine translation tasks compared with strong NMT baselines. We also observe that our method produces meaningful phrases in output languages.
@article{arxiv.1706.05565,
title = {Towards Neural Phrase-based Machine Translation},
author = {Po-Sen Huang and Chong Wang and Sitao Huang and Dengyong Zhou and Li Deng},
journal= {arXiv preprint arXiv:1706.05565},
year = {2018}
}
Comments
in International Conference on Learning Representations (ICLR) 2018