<s>
Interlingual	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
is	O
one	O
of	O
the	O
classic	O
approaches	O
to	O
machine	B-Application
translation	I-Application
.	O
</s>
<s>
Within	O
the	O
rule-based	O
machine	B-Application
translation	I-Application
paradigm	O
,	O
the	O
interlingual	O
approach	O
is	O
an	O
alternative	O
to	O
the	O
direct	B-General_Concept
approach	I-General_Concept
and	O
the	O
transfer	B-General_Concept
approach	I-General_Concept
.	O
</s>
<s>
In	O
the	O
direct	B-General_Concept
approach	I-General_Concept
,	O
words	O
are	O
translated	O
directly	O
without	O
passing	O
through	O
an	O
additional	O
representation	O
.	O
</s>
<s>
In	O
the	O
transfer	B-General_Concept
approach	I-General_Concept
the	O
source	O
language	O
is	O
transformed	O
into	O
an	O
abstract	O
,	O
less	O
language-specific	O
representation	O
.	O
</s>
<s>
The	O
interlingual	O
approach	O
to	O
machine	B-Application
translation	I-Application
has	O
advantages	O
and	O
disadvantages	O
.	O
</s>
<s>
The	O
ideal	O
context	O
for	O
interlingual	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
is	O
thus	O
multilingual	O
machine	B-Application
translation	I-Application
in	O
a	O
very	O
specific	O
domain	O
.	O
</s>
<s>
The	O
first	O
ideas	O
about	O
interlingual	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
appeared	O
in	O
the	O
17th	O
century	O
with	O
Descartes	O
and	O
Leibniz	O
,	O
who	O
came	O
up	O
with	O
theories	O
of	O
how	O
to	O
create	O
dictionaries	O
using	O
universal	O
numerical	O
codes	O
.	O
</s>
<s>
That	O
said	O
,	O
applying	O
the	O
idea	O
of	O
a	O
universal	O
language	O
to	O
machine	B-Application
translation	I-Application
did	O
not	O
appear	O
in	O
any	O
of	O
the	O
first	O
significant	O
approaches	O
.	O
</s>
<s>
However	O
,	O
during	O
the	O
1950s	O
and	O
60s	O
,	O
researchers	O
in	O
Cambridge	O
headed	O
by	O
Margaret	O
Masterman	O
,	O
in	O
Leningrad	B-Device
headed	O
by	O
Nikolai	O
Andreev	O
and	O
in	O
Milan	O
by	O
Silvio	O
Ceccato	O
started	O
work	O
in	O
this	O
area	O
.	O
</s>
<s>
Early	O
interlingual	O
MT	O
systems	O
were	O
also	O
built	O
at	O
Stanford	O
in	O
the	O
1970s	O
by	O
Roger	O
Schank	O
and	O
Yorick	O
Wilks	O
;	O
the	O
former	O
became	O
the	O
basis	O
of	O
a	O
commercial	O
system	O
for	O
the	O
transfer	O
of	O
funds	O
,	O
and	O
the	O
latter	O
's	O
code	O
is	O
preserved	O
at	O
The	O
Computer	O
Museum	O
at	O
Boston	O
as	O
the	O
first	O
interlingual	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
system	O
.	O
</s>
<s>
In	O
the	O
1980s	O
,	O
renewed	O
relevance	O
was	O
given	O
to	O
interlingua-based	O
,	O
and	O
knowledge-based	O
approaches	O
to	O
machine	B-Application
translation	I-Application
in	O
general	O
,	O
with	O
much	O
research	O
going	O
on	O
in	O
the	O
field	O
.	O
</s>
<s>
The	O
most	O
important	O
research	O
of	O
this	O
era	O
was	O
done	O
in	O
distributed	B-Application
language	I-Application
translation	I-Application
(	O
DLT	O
)	O
in	O
Utrecht	O
,	O
which	O
worked	O
with	O
a	O
modified	O
version	O
of	O
Esperanto	O
,	O
and	O
the	O
Fujitsu	O
system	O
in	O
Japan	O
.	O
</s>
<s>
This	O
principle	O
is	O
also	O
used	O
in	O
pivot	O
machine	B-Application
translation	I-Application
,	O
where	O
a	O
natural	O
language	O
is	O
used	O
as	O
a	O
"	O
bridge	O
"	O
between	O
two	O
more	O
distant	O
languages	O
.	O
</s>
<s>
In	O
interlingual	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
systems	O
,	O
there	O
are	O
two	O
monolingual	O
components	O
:	O
the	O
analysis	O
of	O
the	O
source	O
language	O
and	O
the	O
interlingual	O
,	O
and	O
the	O
generation	O
of	O
the	O
interlingua	O
and	O
the	O
target	O
language	O
.	O
</s>
<s>
It	O
is	O
however	O
necessary	O
to	O
distinguish	O
between	O
interlingual	O
systems	O
using	O
only	O
syntactic	O
methods	O
(	O
for	O
example	O
the	O
systems	O
developed	O
in	O
the	O
1970s	O
at	O
the	O
universities	O
of	O
Grenoble	O
and	O
Texas	O
)	O
and	O
those	O
based	O
on	O
artificial	B-Application
intelligence	I-Application
(	O
from	O
1987	O
in	O
Japan	O
and	O
the	O
research	O
at	O
the	O
universities	O
of	O
Southern	O
California	O
and	O
Carnegie	O
Mellon	O
)	O
.	O
</s>
<s>
The	O
following	O
resources	O
are	O
necessary	O
to	O
an	O
interlingual	B-General_Concept
machine	I-General_Concept
translation	I-General_Concept
system	O
:	O
</s>
<s>
One	O
of	O
the	O
problems	O
of	O
knowledge-based	O
machine	B-Application
translation	I-Application
systems	I-Application
is	O
that	O
it	O
becomes	O
impossible	O
to	O
create	O
databases	O
for	O
domains	O
larger	O
than	O
very	O
specific	O
areas	O
.	O
</s>
<s>
One	O
of	O
the	O
main	O
advantages	O
of	O
this	O
strategy	O
is	O
that	O
it	O
provides	O
an	O
economical	O
way	O
to	O
make	O
multilingual	O
translation	B-Application
systems	I-Application
.	O
</s>
<s>
The	O
more	O
languages	O
added	O
to	O
the	O
translation	B-Application
system	I-Application
,	O
and	O
the	O
more	O
different	O
they	O
are	O
,	O
the	O
more	O
potent	O
the	O
interlingua	O
must	O
be	O
to	O
express	O
all	O
possible	O
translation	O
directions	O
.	O
</s>
<s>
Another	O
problem	O
is	O
that	O
it	O
is	O
difficult	O
to	O
extract	O
meaning	O
from	O
texts	O
in	O
the	O
original	O
languages	O
to	O
create	O
the	O
intermediate	B-Application
representation	I-Application
.	O
</s>
<s>
Google	B-Application
Translate	I-Application
use	O
English	O
internally	O
as	O
a	O
pivot	O
language	O
for	O
some	O
language	O
pairs	O
such	O
as	O
Chinese	O
and	O
Japanese	O
,	O
and	O
more	O
generally	O
those	O
with	O
"	O
higher	O
quality	O
"	O
neural-network	O
translators	O
with	O
English	O
but	O
not	O
between	O
each	O
other	O
.	O
</s>
