🚀 go-pugleaf

Xerox currently has something pretty much along the lines of what you are
asking for.   I can create any English document (or a mailnote such as this) -
complete with tables, headers and footers, graphics, etc - and get back an
identical document in a foreign language at almost the click of a button.  All
I do is close the document on my desktop and copy the document icon onto a
globe-shaped software icon.  A short while later, the software icon changes
shape to notify me that my translation is ready.  I open it and copy a new
(translated) document icon onto my desktop.  All the formatting of the original
document is preserved but the text is in one of French, Italian, German,
Spanish, Dutch, Danish, Norwegian, Finnish or Swedish.  Accessible over the
Xerox network, this translation can be achieved from anywhere around the Xerox
world.

We have used this system in Xerox and Rank Xerox translation departments for
the translation of Xerox products for many years.

Sounds good?  Of course, there are a few things to bear in mind.  In
particular:

1.	The software icon strips the text out of the document structure and
sends it to Systran running on an IBM computer in New York.  This is a
'reasonably competent' automatic translation system which Xerox has been
actually using  in a production environment for many years.  NOTE: Systran has
progressed in leaps and bounds since the fairly primitive versions of 10-12
years ago (it seems that most people are unaware of this).  Nevertheless,
machine translation algorithms are still no substitute for a human intelligence
and post editing will be necessary.  How much post-editing depends on the
complexity of the source text and whether or not you want to publish or simply
gain a basic understanding of the foreign text.  Xerox has used Systran
successfully for the translation of service materials for many years.
Basically, the simpler the English, the better the translation.
2.	You need to code entries in a dictionary for the words in your source
text.  This tends to require quite a lot of skill and training.  (The
dictionary allows you to specify lexical, syntactic and  semantic contexts, as
well as targeting specific customers or changing the translation according to
the subject domain.  In addition, you need, of course, to build in all the
necessary 'grammatical' information for each word in your source.)  Note that
there is also a program you can run on your text before translation which
checks that all the simple terms you have used are in the Systran dictionary,
as well as signalling potential problems with long compound nouns, homographs,
etc.

Having said this, there is a commercial version of this software, which I have
to admit I know very little about.  It is called DocuTran.  The brochure I have
in front of me says that it 'runs on Intel PCs and Sun SPARCstations as well as
Xerox Professional Computer Systems.'  It seems to be capable of translating
from French and German into English, as well as English to French, Spanish,
Italian, German and Portuguese.

A final thought:  IF the text you translate is HIGHLY repetitive, using
standard forms or stock sentences, you may get some mileage out of 'repetitions
processing' (example based translation).   I think an interactive CAT system
like ALPs might allow you to do this reasonably quickly.  The system breaks the
text into 'segments' (approximately equivalent to sentences) and searches for
matching source segments in a database.  When a match is found, the translation
associated with the source in the database can be automatically inserted into
the target document.  Of course,  to use this approach successfully, you need a
database of existing translations for all the text you want to translate, and
translations  in each target language required.

Richard E. Birch
----------
Senior Translation Systems Specialist
Customer Business Services, Rank Xerox Technical Centre
Welwyn Garden City, UK.