<s>
Machine	B-Application
translation	I-Application
(	O
MT	O
)	O
algorithms	O
may	O
be	O
classified	O
by	O
their	O
operating	O
principle	O
.	O
</s>
<s>
Rule-based	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
(	O
RBMT	B-General_Concept
)	O
is	O
generated	O
on	O
the	O
basis	O
of	O
morphological	O
,	O
syntactic	O
,	O
and	O
semantic	O
analysis	O
of	O
both	O
the	O
source	O
and	O
the	O
target	O
languages	O
.	O
</s>
<s>
Corpus-based	O
machine	B-Application
translation	I-Application
(	O
CBMT	O
)	O
is	O
generated	O
on	O
the	O
analysis	O
of	O
bilingual	O
text	O
corpora	O
.	O
</s>
<s>
Given	O
large-scale	O
and	O
fine-grained	O
linguistic	O
rules	O
,	O
RBMT	B-General_Concept
systems	O
are	O
capable	O
of	O
producing	O
translations	O
with	O
reasonable	O
quality	O
,	O
but	O
constructing	O
the	O
system	O
is	O
very	O
time-consuming	O
and	O
labor-intensive	O
because	O
such	O
linguistic	O
resources	O
need	O
to	O
be	O
hand-crafted	O
,	O
frequently	O
referred	O
to	O
as	O
knowledge	O
acquisition	O
problem	O
.	O
</s>
<s>
The	O
direct	O
,	O
transfer-based	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
and	O
interlingual	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
methods	O
of	O
machine	B-Application
translation	I-Application
all	O
belong	O
to	O
RBMT	B-General_Concept
but	O
differ	O
in	O
the	O
depth	O
of	O
analysis	O
of	O
the	O
source	O
language	O
and	O
the	O
extent	O
to	O
which	O
they	O
attempt	O
to	O
reach	O
a	O
language-independent	O
representation	O
of	O
meaning	O
or	O
intent	O
between	O
the	O
source	O
and	O
target	O
languages	O
.	O
</s>
<s>
Translation	O
is	O
then	O
performed	O
on	O
representations	O
of	O
the	O
source	O
sentence	O
structure	O
and	O
meaning	O
respectively	O
through	O
syntactic	O
and	O
semantic	O
transfer	B-General_Concept
approaches	I-General_Concept
.	O
</s>
<s>
A	O
transfer-based	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
system	O
involves	O
three	O
stages	O
.	O
</s>
<s>
Statistical	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
(	O
SMT	O
)	O
is	O
generated	O
on	O
the	O
basis	O
of	O
statistical	O
models	O
whose	O
parameters	O
are	O
derived	O
from	O
the	O
analysis	O
of	O
bilingual	O
text	O
corpora	O
.	O
</s>
<s>
Example-based	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
(	O
EBMT	B-General_Concept
)	O
is	O
characterized	O
by	O
its	O
use	O
of	O
bilingual	O
corpus	O
with	O
parallel	O
texts	O
as	O
its	O
main	O
knowledge	O
,	O
in	O
which	O
translation	O
by	O
analogy	O
is	O
the	O
main	O
idea	O
.	O
</s>
<s>
There	O
are	O
four	O
tasks	O
in	O
EBMT	B-General_Concept
:	O
example	O
acquisition	O
,	O
example	O
base	O
and	O
management	O
,	O
example	O
application	O
and	O
synthesis	O
.	O
</s>
<s>
Both	O
belonging	O
to	O
CBMT	O
,	O
sometimes	O
referred	O
to	O
as	O
data-driven	O
MT	O
,	O
EBMT	B-General_Concept
and	O
SMT	O
have	O
something	O
in	O
common	O
which	O
distinguish	O
them	O
from	O
RBMT	B-General_Concept
.	O
</s>
<s>
By	O
contrast	O
,	O
EBMT	B-General_Concept
uses	O
the	O
bitext	O
as	O
its	O
primary	O
data	O
source	O
,	O
in	O
which	O
preprocessing	O
the	O
data	O
is	O
optional	O
and	O
if	O
the	O
input	O
is	O
in	O
the	O
example	O
set	O
,	O
the	O
same	O
translation	O
is	O
to	O
occur	O
.	O
</s>
