1 STATISTICAL MACHINE TRANSLATION MODELFROM A WORD T IN QUESTION Q U...

4.1 Statistical Machine Translation Model

from a word t in question Q using the translation

model. The above formula uses linear interpolation

(f

^(s)

, e

^(s)

), s = 1,...,S is a parallel corpus. In a

smoothing of the document model with the back-

sentence pair (f, e), source language String, f =

ground language model P (t|C). λ is the smoothing

f

₁

f

₂

...f

_J

has J words, and e = e

₁

e

₂

...e

_I

has I word-

parameter. P (t|Q) and P (t|C) are estimated using

s. And alignment a = a

1

a

2

...a

J

represents the map-

the maximum likelihood estimator.

ping information from source language words to tar-

One important consideration is that statistical ma-

get words.

chine translation models first estimate P r(f|e) and

Statistical machine translation models estimate

then calculate P r(e|f) using Bayes’ theorem to min-

P r(f|e), the translation probability from source lan-

imize ordering errors (Brown et al., 1993):

guage string e to target language string f (Och et al.,