Current browse context:
cs.CL
Change to browse by:
References & Citations
Computer Science > Computation and Language
Title: Character-based Neural Machine Translation
(Submitted on 2 Mar 2016 (v1), last revised 30 Jun 2016 (this version, v3))
Abstract: Neural Machine Translation (MT) has reached state-of-the-art results. However, one of the main challenges that neural MT still faces is dealing with very large vocabularies and morphologically rich languages. In this paper, we propose a neural MT system using character-based embeddings in combination with convolutional and highway layers to replace the standard lookup-based word representations. The resulting unlimited-vocabulary and affix-aware source word embeddings are tested in a state-of-the-art neural MT based on an attention-based bidirectional recurrent neural network. The proposed MT scheme provides improved results even when the source language is not morphologically rich. Improvements up to 3 BLEU points are obtained in the German-English WMT task.
Submission history
From: Marta R. Costa-Jussà [view email][v1] Wed, 2 Mar 2016 18:01:57 GMT (76kb,D)
[v2] Thu, 19 May 2016 14:02:48 GMT (77kb,D)
[v3] Thu, 30 Jun 2016 10:28:36 GMT (77kb,D)
Link back to: arXiv, form interface, contact.