Decoding as Continuous Optimization in Neural Machine Translation. (arXiv:1701.02854v1 [cs.CL])

In this work, we propose a novel decoding approach for neural machine translation (NMT) based on continuous optimisation. The resulting optimisation problem can then be tackled using a whole range of continuous optimisation algorithms which have been developed and used in the literature mainly for training. Our approach is general and can be applied to other sequence-to-sequence neural models as well. We make use of this powerful decoding approach to intersect an underlying NMT with a language model, to intersect left-to-right and right-to-left NMT models, and to decode with soft constraints involving coverage and fertility of the source sentence words. The experimental results show the promise of the proposed framework.

from cs.AI updates on arXiv.org http://ift.tt/2j8JFKP
via IFTTT