CAT - Beware of the beast!
Some translators complain that CATs (Computer Aided Translations) tend to decrease quality. While many (dare I say "most"?) of them don't have a clue and confuse CAT tools with machine translation, it is true that there are also quality issues involved with the use of a CAT.
On the one hand, a CAT tool helps increase consistency, (a part of quality), keep the original format, protect code, and handle terminology (and thus help increase quality), but on the other, it has a "darker side" too, and you should know about it.
When you get started with CAT tools and begin to understand what can be done, you get excited - at least, I did. Your productivity increases, translation is easier, faster and it is very tempting to translate everything with your CAT.
That's mostly OK, but somewhere in the back of your mind there should be some awareness that there are documents that are NOT fit be translated with a TM software. (And note that I refer only to the TM engine of a CAT.)
This is my opinion only and is open to discussion.
- Powerpoint presentations
- Communication: A translator translates communication, not text. In a Powerpoint presentation, the layout is a very large part of that communication. You need to see the slides to understand "what" goes "where" and what the intended communication is. With a CAT, you typically translate the text without direct reference to the slide (unless you look at the slide each time you translate a segment - but then you waste a lot of time). This is maybe not true of all CATs, however.
- Segmentation: Sentences are often cut in different text boxes, or with a paragraph mark. Looking at the slide makes it clear enough, but from a segmentation viewpoint, (Segmentation is the heart of a TM engine) it's a mess.
"A great new |...| design"
Typical, but, how do you translate "A great new"? And if you decide to translate by "a great new design" and "design" by nothing (note that you have just altered the design of the slide), your TM will contain the following entries:
"A great new = A great new design" and "design = "
Worthless, not to mention dangerous, because if later on you have "a great new |...| approach"..., your CAT will proudly offer "A great new design" as a 100% match.
- DTP issues: Size. Some languages are typically longer then others. Let's say French is about 10-20% longer than English in average. On a slide, space is limited, and you must see the slide to know just how much space you have. Working with a CAT may mean a LOT of extra DTP work after the translation.
To sum it up, CATs can be used to translate Powerpoint files but overall, I found them to impede productivity, present quality related liabilities, mess up TMs and generate a lot of DTP work.
The exception is when the presentation contains very large notes or large amounts of text... in other words, when the presentation is not really a presentation.
- Ads/Marketing copies
Why? Typical CAT segmentation is sentence-based. This implies that a document can be translated sentence for sentence, from any language to any other.
This works great with technical documents, usually:
"Select the source. Click on whatever. Select the appropriate options and click OK..."
What matters most with a technical document (i.e. a manual) is whether the end user will be able to understand what to do and how. "Style" matters little, beyond clarity and accuracy.
For Ads and Marketing copies however, a sentence by sentence translation will often not do. Style/impact on the target audience is much more important.
Here is a run-of-the-mill example to illustrate that point. (Sorry, French only.)
Welcome to the 2004 XXXXX Dealer Book.
Voici le catalogue fournisseur XXXXX 2004.
In keeping with our 30th Anniversary (1974-2004), this is the most ambitious publication in our company’s history.
Pour célébrer notre 30ème anniversaire (1974-2004), voici la publication la plus ambitieuse de l'histoire de notre société.
Inside, you’ll find complete technical information on the processes, technologies, and products that make the XXXXX brand what it is.
Vous trouverez à l'intérieur des informations techniques complètes sur les processus, les technologies et les produits qui ont fait de la marque XXXXX ce qu'elle est devenue.
And here is the result:
Voici le catalogue fournisseur XXXXX 2004. Pour célébrer notre 30ème anniversaire (1974-2004), voici la publication la plus ambitieuse de l'histoire de notre société. Vous trouverez à l'intérieur des informations techniques complètes sur les processus, les technologies et les produits qui ont fait de la marque XXXXX ce qu'elle est devenue.
It is not "bad", but it feels like a translation. It doesn't flow, which is a logical consequence of sentence-level segmentation.
Welcome to the 2004 XXXXX Dealer Book. In keeping with our 30th Anniversary (1974-2004), this is the most ambitious publication in our company’s history. Inside, you’ll find complete technical information on the processes, technologies, and products that make the XXXXX brand what it is.
En célébration de notre 30ème anniversaire, nous sommes heureux de vous présenter le catalogue fournisseur XXXXX 2004, la publication la plus ambitieuse dans l'histoire de notre société. Vous trouverez à l'intérieur des informations techniques complètes sur les processus, les technologies et les produits qui ont fait de XXXXX un leader incontesté sur le marché de YYYYY.
This second translation is a lot more natural. It doesn't feel so much like a translation no more. The phrase in green is not part of the original paragraph but from another paragraph in the document. In French however, it fits better here. That's what the author meant.
Ad/marketing copies require a global understanding of the communication rather then a sentence by sentence approach, imposed by classic segmentation. In other words, using a TM will tend to reduce your overall quality level by focusing your attention on sentences instead of the whole text.
You could indeed decide to segment at paragraph level, but then lose most of the leverage from the TM (What are the odds of a full paragraph repeating, even fuzzily?) and the segments would often be so big that you would often face display issues.
A CAT can nevertheless help by handling terminology, to executing concordance searches, accessing dictionaries,... but using the TM as you usually do could reduce the quality of your output
Theses are, in my opinion, the main drawbacks of using a TM engine. I may write more on the issue later. I would be very interested to hear your opinions on that subject, especially if you have other related issues to present.