Working with CAT tools

CAT – Computer Assisted Translation, is not the same as machine translation. Working with CAT tools cannot be compared to using a program such as Google Translate. With Google Translate you feed in your source text and the program handles the translation for you. However, machine translation software often produces these hilarious mistakes because the program does not understand that identical words mean different things in different contexts. A famous howler is "Come alive with the Pepsi Generation", which the company translated into "Pepsi will bring your ancestors back from the dead."

In contrast, CAT tools are software that stores your own translations in its memory, so if you get more work on the same or a very similar project, the CAT tool will retrieve those translations for you. Also, it recognises repetitions within a text and automatically inserts the translations.

CAT tools greatly enhance productivity and ensure consistent translations.

They are particularly well suited to technical and repetitive texts, such as manuals, handbooks, product descriptions. They don't work quite as well with literary texts, the translation of novels and poetry.  But for most translators, they are part of their daily work.

CAT tools can be extremely helpful, because they are capable of showing the translator their own previous translations and, what is more, automatically inserting translations into repeated sentences.

But it does give rise to a huge question in the translating industry: What to charge for repetitions and matches. Matches are instances where a translation memory (TM) provided by the agency, already contains the translation of a sentence. Therefore the translator simply has to accept that translation. What is more, matches are expressed in percentage points, from 100% the same as the sentence stored in the memory down to 0% the same.

Agency clients in particular use CAT tools and expect freelance translators to charge differentiated rates. They neither want to pay for 100% repeated text nor for 100% matches, and they want to pay less than full price for 70-99% matches.

It makes sense. If you don’t have to translate it, or only make a minor change, say from red skirt to red trousers, why should you get paid as much for that as if you had to translate the complete paragraph?

And yet, for translators, it’s not quite that simple.

A German language example of working with CAT tools

Not everything that is a 100% repetition in one language is also a 100% repetition another. Sometimes, one word has several meanings and just happens to have the same spelling.

And so, in a recent tourism text on various winter sporting events and competitions in South Tyrol, there was a frequent one liner: "Preise:".

In most cases, it meant prices. But on two occasions it meant prizes.

So if your text is about a sporting event at which prizes are given out, but if there is an entry price for spectators, your CAT tool segment Preise can have many 100% repetitions of Preise. But you still have to check whether any repetition refers to prices or prizes.

If clients are not willing to pay translators anything at all for any repetitions, they will not even look at them. They will filter them before they start the translation and lock those sentences, so they don’t even appear on the screen. As a consequence, they would not spot instances such as the above one.

Matches below 100% are even more complicated. You’ve changed "red skirt" to "red trousers". Because the change only affects two words in English, the client only wants to pay a fraction of the standard price for the sentences concerned, .

While that is true in English, it is not true in German. Here, a skirt is grammatically an it, while trousers are feminine, she.  And that impacts not only on the translation of red and green, but also on the translation of other words in that sentence. It is therefore likely that the translator has to change several words have, including the word order in a sentence, when all you do in English is change two words.

An English language example

An English language example for the effects of working with CAT tools would be substituting  "umbrella" for "raincoat". It’s "a" raincoat but "an" umbrella. So if you change raincoat to umbrella, you have to also change any a’s in front of the word, or you end up with "a umbrella".

A particular challenge are questionnaires like this:

What is the first thing you do when you get up in the morning?

  1.        clean your teeth
  2.        make a cup of tea
  3.        feed the dog

Your next question could be:

What do you say most frequently to your child in the morning?

  1.        clean your teeth
  2.        make a cup of tea
  3.        feed the dog

Exactly the same 3 answers in English. So they’re 100% repetitions.

Only, again, in German they’re not.

The first set of three are plain statements of what you do, the second set of three are requests. In German, they are two completely different translations. "Clean your teeth" changes from "die Zähne putzen" to "putz dir die Zähne".

When your translator isn’t immediately willing to give you massive discounts for repetitions and matches, please don’t assume that they are trying to rip you off. When working with CAT tools, translators are usually willing to consider giving discounts. But only, if they are 100% sure that clients' repetitions are repetitions and that what they think are matches are matches in both languages. Otherwise it pays to be careful.