Enabling Multi-Source Neural Machine Translation By Concatenating Source Sentences In Multiple Languages

  • 2019-03-04 04:10:10
  • Raj Dabre, Fabien Cromieres, Sadao Kurohashi
  • 0


In this paper, we explore a simple solution to "Multi-Source Neural MachineTranslation" (MSNMT) which only relies on preprocessing a N-way multilingualcorpus without modifying the Neural Machine Translation (NMT) architecture ortraining procedure. We simply concatenate the source sentences to form a singlelong multi-source input sentence while keeping the target side sentence as itis and train an NMT system using this preprocessed corpus. We evaluate ourmethod in resource poor as well as resource rich settings and show itseffectiveness (up to 4 BLEU using 2 source languages and up to 6 BLEU using 5source languages). We also compare against existing methods for MSNMT and showthat our solution gives competitive results despite its simplicity. We alsoprovide some insights on how the NMT system leverages multilingual informationin such a scenario by visualizing attention.


Introduction (beta)



Conclusion (beta)