Bridging the inflection morphology gap for arabic statistical machine translation

Statistical machine translation (SMT) is
based on the ability to effectively learn
word and phrase relationships from par-
allel corpora, a process which is consid-
erably more difficult when the extent of
morphological expression differs signifi-
cantly across the source and target lan-
guages. We present techniques that se-
lect appropriate word segmentations in
the morphologically rich source language
based on contextual relationships in the
target language. Our results take ad-
vantage of existing word level morpho-
logical analysis components to improve
translation quality above state-of-the-art
on a limited-data Arabic to English speech
translation task.