rail transport l services l production support
of Euros every year, and the time required
for the completion of different language
versions of manuals has also been significantly reduced. Let us look at the basic
principles of CAT software, which significantly supports the effectiveness and quality of the localisation process.
First of all, it should be emphasised that
CAT software does not replace the work
of the translator, it only makes their work
as efficient and precise as possible. It uses
a translation memory, allowing the software to remember every previously translated sentence. If an identical sentence
occurs in the future, it will be automatically pre-translated (obviously with the possibility of proofreading by the translator).
With this function, significant time and financial savings can be achieved for texts
where some passages often repeat (typically in technical documentation). Another advantage of this functionality is consistency of translation and terminology
throughout the documentation, which
is very desirable for the overall quality of
If the software does not find a 100%
identical sentence in the translation memory, it can find a sufficiently similar sentence instead, i.e. the "fuzzy logic" is applied. For example, if only one or two
words have changed in a long sentence
while the rest of the sentence is identical.
In this case, the translator is offered a "fuzzy match", in which they can immediately
see colour-highlighted differences in both
of the sentences. Then, they can accept
the proposed translation and only translate the different words. Again, this leads
to higher efficiency and consistency of translation throughout the documentation, i.e.
more effective and higher quality translation work. The aim is to re-use the translator's work using a sophisticated comparison algorithm, achieving typically 30-70%
savings in the time and costs involved in
technical documentation translations.
Translation quality is also supported by
a feature for maintaining terminology consistency, which is especially important for
We should also mention a recent, rapidly developing technology. Everyone has
heard of Machine Translation (MT), or has
at least tried a free online translator.
These often produce funny translations
and cannot be relied on if precise wording
is important, which is obviously required
in the case of technical documentation,
contracts, or marketing materials. Therefore, translators need not worry about losing their jobs; however, the way they
work is gradually changing. After two decades of using translation memory and
fuzzy logic, specialised machine translation systems tailored to a specific subject
area and a specific client's needs are becoming increasingly common. We have
been developing the STAR MT machine
translation statistical system for several years in cooperation with university professors of computational linguistics. For our
clients' large and long-term projects, we
prepare individual algorithms from bilingual and monolingual corpora and client
terminology. The result is a very effective
solution, where proposals from the translation memory and fuzzy matches are
supplemented with machine translation
proposals which the translator checks
and, if necessary, modifies. Our client thus
achieves significant time and cost savings
while maintaining the high quality of documentation translations.
For the MT technology to be effective,
the project should have a certain size and
continuity: at least around 500 to 1,000
pages of new text per year. The content
of the text should be simple, factual, and
instructional, e.g. user manuals or technical documentation, intended for accurate
and direct translation, instead of literary
and creative. The original text should be
of a good quality: with uniform terminology, formatting, style, abbreviations, etc.
We expect a literal and direct translation
from the machine translation. For this
technology to save time and costs, we
cannot expect the translator to rearrange
texts or significantly change sentence
structures when editing the machine
translation proposals. Not all language
combinations provide identical results.
Romance and Germanic languages provide good results, and so do some Slavic
languages. However, forget about Hungarian or Estonian, the MT systems cannot
comprehend the complexity of these languages yet. The use of machine translation should be one of the intermediaries
of the whole process, from the efficient
and high-quality creation of source text
through the use of translation memory,
fuzzy matches, and terminology, up to the
application of machine translation and its
regular improvement by importing new
texts into the MT algorithm.
Another emerging trend is the optimisation of technical documentation preparation. There are cutting-edge software tools which help the technical
documentation author to formulate the information presented, e.g. in a technical
manual, as efficiently as possible. Based
on the sentence being written, the smart
software proposes similar texts from previous documentation, on which the author can draw. In addition, a notification
function for forbidden terminology is available, proposing a suitable synonym. All
of these functionalities, including morphological analysis, are also available for
documentation authors in the Czech language. The result is documentation with
a uniform style and terminology, the subsequent translation of which into the target languages is faster and cheaper, as the
proposed texts already have their translation in the translation memory.
Whether you need a translation for your
products as an exporter, or for products
imported from abroad, you can save a lot
of costs and time by streamlining the localisation process. Also, in the Czech Republic, you can obtain CAT software and
tools for optimising the preparation of documents for your documentation department and have your in-house translators
and authors trained. When outsourcing
translations, choose an agency which uses
CAT software and is willing to share its
financial benefits with you.
Ing. László Jankovics
tel.: +420 608 320 100