|
tranfree issue 11 - 15 February 2000
Translation Technology
By Bert Esselink
This is my second contribution in a series of four articles about
software localisation. The previous article discussed the
differences between translation and localisation, and below I
will cover the types of translation tools that are used by
translators working for localisation service providers.
First of all, a distinction needs to be made between machine
translation (MT) tools and computer aided translation (CAT)
tools. Where machine translation tries to replace a translator to
a certain extent, computer aided translation tools support the
translator by preventing repetitive work, automating terminology
lookup activities, and recycling previously translated texts.
Machine translation has not been applied much in the software
localisation industry mainly because, unlike in the automotive
and aerospace industries, software publishers never really
created their documentation in a structured way that would make
machine translation successful. Even though this seems to be
gradually changing, I will focus on computer aided translation
tools in order to reflect current practices in the localization
industry.
Computer aided translation tools, also called computer assisted
translation tools, can be categorized as follows...
- Translation Memory tools
- Terminology tools
- Software Localization tools
The first two types, translation memory and terminology tools,
are typically combined for translation of documentation, online
help, or html text. Software localization tools are used to
translate and test software user interfaces, i.e. dialog boxes,
menus, and messages.
Translation Memory tools
Basically, a translation memory system is no more than a database
which stores translated sentences.
When a source text is imported into a translation memory tool,
the text is segmented. Usually segmentation is performed at
sentence level, where segments are separated by colons, commas,
question marks, etc. However, it is also possible to segment
texts on a paragraph basis, where segments are separated by
paragraph marks. Each segment is a record in the translation
memory database, and each record can store several fields, such
as source text segment, translated segment, language, name of
translator, date of translation, or type of text. The number of
possible data fields in records differs per translation memory
tool.
When text segmented by a translation memory tool is translated,
all translations are automatically stored in the records
containing the source segments. If identical or similar sentences
occur in the source text, the translations are automatically...
 ...
retrieved from the database and inserted into the target text. An
identical segment that is automatically translated is called a
full match; a similar sentence that is automatically translated
is called a fuzzy match. Obviously, fuzzy matches need to be
post-edited to make them correspond to the source text. A fuzzy
match is for example a sentence where only one word has changed.
On large projects, translation memory databases can be shared
amongst a team of translators. This means that if translator A
has translated a sentence which also occurs in the text that
translator B is translating, A's translation will automatically
be retrieved from the translation memory database and inserted in
B's target text.
Naturally, translation memory tools are especially useful on
large volume texts, which contain a lot of repetitive text and
where translations can be created on a one-to-one sentence basis.
Using translation memory tools to translate marketing text or
adverts is not a good idea, simply because those types of texts
often require many adjustments, rewrites, and other modifications.
In the software localisation industry, translation memory tools
have always been very popular because of the short life cycle of
software products. Most software products are updated at least
once a year, and re-using translations of previous versions will
decrease time to market of localised versions drastically.
Examples of translation memory tools are Trados Translator's
Workbench (www.trados.com), Atril Déjà Vu (www.atril.com), and
STAR Transit (www.star-ag.ch).
Terminology tools
In localisation, terminology management is usually dealt with in
a very basic manner. Localisers typically don't create or use
large multilingual terminology databases with term definitions,
context, grammatical information, source, etc. Instead, in most
cases only bilingual glossaries of translated terms or phrases
are used, for example all translated terms from the software user
interface, e.g. menu options, dialog box items, etc. For this
reason, not only are professional terminology management tools
used, but also basic glossary tools with search features.
Most translation memory tools come with terminology management
applications which can be linked to the translation memory for
automatic terminology lookup. Automatic terminology lookup means
that terms in the source text which are found in the dictionary
or terminology database are automatically displayed with their
translations.
Examples of terminology tools are Trados Multiterm
(www.trados.com), Atril TermWatch (www.atril.com) or STAR
TermStar (www.star-ag.ch).
Software Localisation tools
Special tools have been developed to translate graphical user
interfaces of software applications, i.e. the dialog boxes,
menus, and messages that are displayed on a computer screen.
These tools allow translators to view their translations in
context. For example translations can be entered directly in a
dialog box and then saved.
Software localisation tools also contain features for
automatically translating updated software with previously
translated versions, and for running basic tests on localized
software, for example checking if no translated text has been
truncated in the screens because of space restrictions.
Examples of software localisation tools are Corel Catalyst
(catalyst.corel.ie) and Passolo (www.passolo.com).
The Next Generation
Even though many translators still need to get...
 ...
acquainted with traditional translation technology such as translation memory,
the next generation translation tools have already been
introduced. Companies like e-Translate and Uniscape offer
automated internet-based translation workflow solutions that
automate many steps in translation projects. Texts are not only
transferred automatically through each translation and review
phase, but databases containing the source text are linked to
translation technology that detects changed content and then
first pre-translates it using a combination of translation memory
and machine translation before it is forwarded to a human
translator for post-editing.
These so-called translation portals and internet-based
translation management solutions are especially designed to deal
with frequently changing content, such as text published on daily
updated web sites. For more information about these types of
translation technology, visit www.uniscape.com or
www.e-translate.com.
Technology and the web will totally change the way translations
have been done for many centuries. Frequently updated content,
geographically distributed resources, and pressure to keep prices
down will result in further integration of technology and
workflow automation in translation processes. Undoubtedly the
tools can be developed quickly, but the real challenge will be to
keep producing quality translations and to still enjoy
translating.
Bert Esselink has been active in localization since 1990.
After graduating in technical translation and doing coursework in
programming and computational linguistics, he worked for several
years as a software translator, localization engineer, and
technical manager. In 1996 he joined ALPNET in Amsterdam as
localization manager, training new engineers and localization
specialists and coordinating software localization projects and
since early 1999 as globalization manager developing production
standards. As of 1 January 2000 he is employed by INT'L.com
in the Netherlands, managing consultancy services. His first
book, A Practical Guide to Software Localization, was published
in 1998 by John Benjamins Publishing Company (www.benjamins.com). The
second edition is now out and further information is available at
http://www.locguide.com/
Click here to read the next article
Click here to return to tranfree
11 main page
| 