The objective of any translation is to move an original concept into a new environment. This may be e.g. the relatively linear translation of a technical text into another language, the challenge of finding equivalent academic terminology to describe a foreign degree, or the associative task of localizing an advertising text.
In over 20 years of translating technical and publicity texts for world leaders in the process industry and business consulting sectors and a variety of public research institutions, I've encountered a highly diverse cross-section of such environments. Technical, legal and financial concepts require the most direct possible translation, with no loss or addition of meaning. Translating technical sales and strategy documents requires accuracy as well as a certain ease of expression and accessibility. Press, public relations and advertising materials call for translation of style as much as content.
Finding the right combination of precision, accessibility and appropriateness of style is the key to a successful translation. Any one of these dimensions may take precedence in any given translation: The voice of an employment contract should not be playful, a poster headline should not be dull, and a project proposal has to maintain its focus, regardless of the language in which it's written.
The Allure of Machine Translation
Technology is always attractive and rarely, if ever, completely neutral. Translation tools working at the syntactic level may ensure consistency, but also encourage translation by closest possible match, rather than by supporting the proper linguistic process in the target language.
It is not always efficient to cut translation time by postponing the issue of quality to a painstaking post-editing phase. Unless, of course, you're building a bookcase: A 90 degree angle looks much the same in any language.
The CLS Solution
This is why I developed and use a relational database solution which provides support primarily at the lexical level. Interfaces to order management and customer databases instantaneously return not only lexical information (how can this term be translated?) but also information on context, customer and author (where was this particular translation used, when and for whom?). Incorporating data collected over almost 25 years, the CLS system contains more than 40,000 entries, indexing well over 150,000 instances of specific use.
Consistent terminology usage is guaranteed, but the target text doesn't taste metallic, as if mapped onto a list of the most convenient available sentences already used for another purpose.
The art of acquiring the voice of the original concept is thus left with the translator, where it belongs.