Computer-assisted translation

Catalogues, manuals, forms, financial statements and audit reports are all examples of texts which are both complicated and repetitive. Additionally, they often involve the tedious task of copying names, figures and complicated codes. Luckily, tools exist which allow translators to avoid this and having to remember how they previously translated an identical phrase.

Let’s get one thing straight: we are not talking about automatic translation (or AT). This involves a program into which an original text is fed and comes out translated. We use the term “translated” in its loosest possible sense, since what comes out of such a process is often so unintelligible that it cannot even be corrected.

Instruction manuals are the easiest texts for this type of AT software to translate, as they are usually simple and made up of short, clearly divided sentences which include subject, verb and object. Little interpretation is required for the machine to produce a translation, although it can often slip up by choosing the wrong translation of a polysemic term. Although there are some fairly advanced programs on the market, they still cannot take the place of a professional translator when it comes to understanding (and, often as not, correcting) the original.

Translation is often defined as the process of converting, from one language to another, the content or message of the original, respecting or adapting the style where appropriate. Automatic translation programs not only are incapable of understanding a message, but also never reflect its style: it is thus impossible for them to do a proper job.

In contrast, translation memory (TM) software is used to automate certain parts only of the translation process, and is thus more useful, since it leaves the creative process in the hands of a human translator, merely reminding him or her how an identical (or similar) paragraph has been translated in the past.

With TM, the translator overwrites the original sentence. This process avoids copying out numbers, names and complicated codes. TM then memorises the sentence to be able to “recall” it later on in the text. When the translator reaches an identical (or similar) sentence, the software immediately proposes the translation validated earlier by the translator. A good TM program can even adapt the proposed translation, when the differences are limited to numbers or names. The translator must always give his or her approval to these suggestions, but obviously avoids having to re-write the same sentence.

The extra work involved in creating the translation memory (i.e. the first translation with a TM program) is compensated during later updates to the document. However, it is also true that working with these programs, creating a memoryfrom the scratch, is a hard, slow task. Given this, TM can be viewed as a long-term investment, especially suited to catalogues and manuals, which are often highly repetitive and subject to periodic updating.

The most common programs of this type are Trados, Transit, and Déjà Vu. We at eurolink use Déjà Vu because itis the most powerful and versatile program of its type, able to recognise the latest versions of common programs and capable of working with the formats of all other M.T. programs.