Abstract:
Word reordering is a problematic issue for
language pairs with significantly different word orders,
such as the translation between a subject-verb-object
(SVO) language and a subject-object-verb (SOV)
language. When translating between language pairs
with high disparity in word order, reordering is
extremely desirable for translation accuracy. In this
paper, the future research directions of reordering
models for Myanmar-English statistical machine
translation (SMT) are also depicted. In this reordering
model, the word order on source-side is arranged into
the target side word order, before SMT system is
applied. We propose the use of recurrent neural
networks (RNNs) to model preordering for SMT.
