Towards String-to-Tree Neural Machine Translation

https://arxiv.org/abs/1704.04743

Roee Aharoni, Yoav Goldberg

(Submitted on 16 Apr 2017)

We present a simple method to incorporate syntactic information about the target language in a neural machine translation system by translating into linearized, lexicalized constituency trees. An experiment on the WMT16 German-English news translation task resulted in an improved BLEU score when compared to a syntax-agnostic NMT baseline trained on the same dataset. An analysis of the translations from the syntax-aware system shows that it performs more reordering during translation in comparison to the baseline. A small-scale human evaluation also showed an advantage to the syntax-aware system.

Comments:Accepted as a short paper in ACL 2017Subjects:Computation and Language (cs.CL)Cite as:arXiv:1704.04743 [cs.CL] (or arXiv:1704.04743v1 [cs.CL] for this version)

Submission history

From: Roee Aharoni [view email] 
[v1] Sun, 16 Apr 2017 09:54:50 GMT (637kb,D)