Multi-Source Syntactic Neural Machine Translation
This work addresses translation quality for NLP practitioners, offering a method that maintains performance without parsed input and on long sentences, though it is incremental as it builds on existing parsed and seq2seq approaches.
The paper tackles the problem of incorporating source syntax into neural machine translation by introducing a multi-source technique using linearized parses, resulting in an improvement of over 1 BLEU on the WMT17 English-German task compared to baselines.
We introduce a novel multi-source technique for incorporating source syntax into neural machine translation using linearized parses. This is achieved by employing separate encoders for the sequential and parsed versions of the same source sentence; the resulting representations are then combined using a hierarchical attention mechanism. The proposed model improves over both seq2seq and parsed baselines by over 1 BLEU on the WMT17 English-German task. Further analysis shows that our multi-source syntactic model is able to translate successfully without any parsed input, unlike standard parsed methods. In addition, performance does not deteriorate as much on long sentences as for the baselines.