We propose a novel dependency-based reordering
model for hierarchical SMT that
predicts the translation order of two types
of pairs of constituents of the source tree:
head-dependent and dependent-dependent.
Our model uses the dependency structure
of the source sentence to capture
the medium- and long-distance reorderings
between these pairs of constituents.
We describe our reordering model in detail
and then apply it to a language pair in
which the languages involved follow different
word order patterns, English (SVO)
and Farsi (free word order being SOV the
most frequent pattern). Our model outperforms
a baseline (standard hierarchical
SMT) by 0.78 BLEU points absolute, statistically
significant at p = 0.01.