A94-1016 |
Nirenburg 1 INTRODUCTION Current
|
MT systems
|
, whatever translation method
|
A00-1004 |
will compare it with two existing
|
MT systems
|
. Both systems were tested in
|
A94-1016 |
- ture . Since the individual
|
MT systems
|
operate completely independently
|
A00-2026 |
Langkilde and Knight , 1998 ) . The
|
MT systems
|
of ( Berger et al. , 1996 ) learn
|
A97-1048 |
to judge the quality of current
|
MT systems
|
, but they may need help from
|
A00-1004 |
with the same quality produced by
|
MT systems
|
. <title> PartsID : A Dialogue-Based
|
A94-1016 |
uses output from a variety of
|
MT systems
|
working on the same text . In
|
A00-2026 |
comparison suggests that statistical
|
MT systems
|
could also generate text from
|
A97-1048 |
interactive support for conventional
|
MT systems
|
does n't seem suitable for these
|
A00-1004 |
translation model . 4.2 Comparison With
|
MT Systems
|
One advantage of a parallel text-based
|
A94-1007 |
many possible structures that
|
MT systems
|
can not select the correct one
|
A94-1016 |
best results from a variety of
|
MT systems
|
working simultaneously on the
|
A94-1043 |
machine acceptable sentences ,
|
MT systems
|
for bilingual and monolingual
|
A94-1007 |
many possible structures that
|
MT systems
|
can not select the correct one
|
A94-1007 |
possible structures are produced that
|
MT systems
|
can not select the correct one
|
A83-1028 |
spectrum -- around 90 % -- for
|
MT systems
|
that vary remarkably v.r.t. the
|
A94-1036 |
memory systems cost a fraction of
|
MT systems
|
) or they are simply inappropriate
|
A97-1047 |
To avoid these problems , the
|
MT systems
|
developed recently generally
|
A83-1026 |
design . In large general purpose
|
MT systems
|
, it is necessary to assume that
|
A00-1002 |
into the development of various
|
MT systems
|
have been wasted and have not
|