[ Pobierz całość w formacie PDF ]

80 HOME OFFICE SETUP
directly into a word processing program. On projects like this,
speech recognition software can help you work faster and with
less strain to your hands and arms.
3.8 Translation memory software
Probably one of the most frequent topics of conversation among
translators is whether to purchase translation memory software,
which software works best for a particular application, how much
the software costs, and on and on and on. Conceptually, transla-
tion memory software is not very complex. The most important
thing to understand is the difference between translation mem-
ory software (TM; sometimes also referred to as computer-assisted
translation software, or CAT) and machine translation software (MT).
Translation memory software doesn't do the translation for you,
rather it helps human translators work faster and more accurately
by recycling material that has already been translated and sug-
gesting a match between the old translation and the current one.
Machine translation software is translation done entirely by a
computer. Machine translation is currently the subject of a great
deal of research and development, and it sometimes works well
enough to get the basic idea of a document, but often produces
comical or totally incomprehensible translations otherwise.
By definition, translation memory software only works with
electronic documents; you can't take a piece of paper and run it
through a translation memory program unless you retype it or
scan it first using optical character recognition software (OCR),
you translate mostly from hard copy or scanned documents,
so if
translation memory software is not very helpful. However, most
translation memory programs can pull the text out of spread-
sheets, HTML files, etc. Translation memory software works by
segmenting your source document, meaning that the program
breaks your document up into smaller chunks, normally sen-
tences but sometimes paragraphs. Then, when a segment is ready
to be translated, the program checks to see if you already trans-
lated a similar segment, and if it finds a match it suggests the
TRANSLATION MEMORY SOFTWARE 81
match to you, theoretically resulting in a faster and more con-
sistent translation. For example, if you already translated the
sentence "This is a cat," and the next sentence was "This is a black
cat," the TM program would suggest "This is a cat" as a match, so
that you only had to type "black" in the target text box, instead of
typing the whole sentence. Most TM programs display potential
matches as percentages, for example the sentence "This is a dog"
would be a 75% match with the sentence "This is a cat," since
only one of the four words is different. This matching feature can
be particularly helpful when your translation client has specific
terms that they want you to use throughout the document, for ex-
ample to always use "President and Chief Executive Officer" for
the chairman of the company. Most translation memory packages
use the file terms uncleaned and cleaned; an uncleaned file contains
the source text and the electronic codes used by the translation
memory program, while the cleaned file contains only the target
text.
Another use of translation memory tools, although a task that
some TM tools don't do very well, is
alignment. Alignment means
taking the source and target versions of a document, and matching
them up so that you have pairs of sentences, one in the source
language and one in the target language. This way, you can
create a bilingual glossary out of your old translated documents.
In practice, this function can be annoying to use; if the source
and target sentence pairs don't match up exactly, it requires a lot
of time on the translator's part to manually fix the mis-aligned
segments.
Translation memory software is also somewhat controversial
among translators. One of the reasons for this is that translation
clients who are aware of TM software's capabilities will often
ask for discounts on repetitive documents; for example the client
will use the software to analyze a document, and tell you that
although the document is 2,782 words, they only want to pay
for 2,582, because there are 200 words that are repeated in the
document. Or, a client might ask you to reuse the translation
memory file from an old translation, and want to pay only for the
new words translated; for example if the client is putting out a
82 HOME OFFICE SETUP
new version of a software manual, they might want to pay you to
translate only the updated parts. Some translators are completely
opposed to giving discounts for the use of TM software, on the
grounds that they pay to acquire and maintain the software, they
do the work on the translation, and even if a segment is a 100%
match with a previously translated one, the translator still has
to read the segment and sometimes make other adjustments as [ Pobierz całość w formacie PDF ]