| Traditional "human"
translation
Once the product-specific terminology is set, the translator or
agency is ready to work on the translation of the text to produce
a clear and accurate document in the foreign language. In addition
to the linguistic aspects, other important details must be handled,
particularly the conversion of units of measure. Then the translation
must be edited and reviewed to check style, completeness, and accuracy
of content.
As noted above, the expert translator will have technical experience,
but every client will use its own particular "dialect."
The translator should continue to have access to a contact person
in the company for ongoing help with defining any unfamiliar terms
and clarifying usages that are unique to the company. As more and
more documentation is translated, the terminology becomes well-defined
and this stage goes more quickly.
One advantage of using an agency is the data-basing of terminology
by client and by product (see Glossary-Building).
In that case, if a specific translator is no longer available, the
terminology and the previous translations are stored for others
to use.
Machine translation
If you have large-scale, long-term translation needs, machine translation
may be a viable option. The number of affordable packages that run on the
PC is increasing much more quickly than the general public's ability to
evaluate their usefulness. For this reason, a short section on machine
translation is given here to provide some background information.
What is machine translation? Machine translation
is the use of computer programs to translate or help translate materials
from one language into another. Available programs cover the spectrum
from PC packages priced at $19.95 to workstation programs that cost
thousands of dollars.
Generally speaking, you get what you pay for in machine translation.
The inexpensive programs are based on simpler technology, the more expensive
are based on more complex technology. Some of the biggest players on the
high end are Systran, Logos, and Intergraph.
Translation with the help of computers is usually divided into two categories:
true machine translation (MT) and computer-aided translation (CAT).
- (True) machine translation means
that the actual translation effort is performed by a computer
program, i.e., the computer will take any text in a source language
and attempt to translate it into a target language.
- Computer-aided translation describes a process
in which a computer program assists in the translation effort
using glossaries, matching, and other interactive aids, but does
not perform the actual translation.
Are translators needed with
machine translation?
Although MT and CAT are becoming more sophisticated every
year, they still require human intervention on several levels. In other
words, they can speed up the translation process, but they cannot replace
the translator (all the systems in use today require some level of post-editing
of the computer output).
Some of the lower-priced packages can be used to translate light correspondence.
The output may not be perfect, but you can probably get the sense of the
text.
The higher-end packages require extensive programming, especially at the early
stages. You will need to create and
input a dictionary of your technical terminology, so that RAM,
for instance, will be translated as "random access memory,"
and not as "male sheep."
The translated files output from the program will require post-editing.
This is the process whereby a translator/linguist compares the output
translation to the source document and corrects the translation.
The cost-effectiveness of the post-editing is directly proportional
to a) the quality of the program, b) the quality of
the writing in the source document (see "Preparing for Translation"),
and c) the amount and quality of the dictionary work. If
properly developed and maintained, it may be possible to double
or even triple output per translator.
As a general rule, high-end MT is not an effective solution for
small volumes of translation (the cost is simply not justified).
However, it can be a long-term solution to high-volume translation
needs, with the caveat that the requisite investment must
be made up front in order for it to pay off. If inadequate attention
is paid to maintaining your specific dictionaries and writing the
source document for translation, then the computer output is unlikely
to rise above a low level of quality, and there will be no time-saving
benefit.
What are some MT/CAT strategies
used today?
The most frequent are:
- contracting the entire MT process (dictionary-building, maintenance,
post-editing, etc.) through an agency that provides the service;
- purchasing and maintaining an MT system on-site, and contracting
out post-editing;
- purchase, setup, and maintenance of the entire MT and post-editing
process on-site. The level of on-site investment is lowest for the first
item and highest for the last. The most appropriate approach will depend
on your company's specific needs and situation.
The rise of translation memory tools
The past few years have seen an increase in the use of translation memory applications
as a special form of CAT tools. Translation memory tools are database
applications which compare a source text to a database of matching
source and target strings. In its pure form, this approach does
not perform an independent assembly of translated phrases based
on syntax and semantics rules; instead, it relies heavily on fuzzy-match
logic. Translation is performed in an interactive process between
human translator and computer.
This technology is comparatively inexpensive, relatively easy to set up and use, and
suitable to projects of medium as well as large size. It is especially useful for
projects which undergo many revision cycles.
|
