CAT software is an essential tool for any professional translator looking to maintain a competitive edge. “CAT” stands for Computer Assisted Translation; essentially, CAT software will take any standard file containing text, and allow a translator to go through the text piece-by-piece with the assistance of various computerised translation tools. It boosts work rate significantly, and in some cases can also improve the quality of the translation produced. This applies to almost any kind of text; possibly in the case of poetry it might not offer a great deal of assistance, but for any other text, CAT software, when used properly, will improve the quality of the translation.
“When used properly” is an important caveat, however. There are a number of pitfalls related to overreliance on the software, and it is important not to expect it to translate the text for you, but to think of it as a tool for arranging the text so that it becomes easier to translate.
In this post I will describe the three main features of CAT software, segmentation, translation memory and machine translation, and discuss both how they help the translator and the pitfalls that can be caused by overreliance.
Segmentation might seem like the least important of the three main features of CAT software, but it makes a significant contribution to the speed at which the translation is carried out. By segmentation, I mean that the software divides the text to be translated into segments – usually, but not always, sentences – for the translator to translate one by one. By dividing the text up into digestible chunks this way, the software prevents the translator from getting bogged down in huge paragraphs, instead enabling him or her to progress smoothly and methodically through the text. Even without the other two features, I find that this feature allows me to translate significantly faster.
However, one does need to be a little wary of overreliance on the segmentation feature. The disadvantage of segmentation is that it can lead the translator to translate each segment without considering its context, leading to a disjointed translation that does not effectively communicate the style and the sense of the original text. To avoid this pitfall, it is necessary to both consider the context carefully at the moment of translation and to proof-read the whole text after the CAT translation alongside the original to ensure that it is a consistent and faithful translation.
Translation memory is a feature of CAT software which remembers how you translate a specific word or phrase and, if it recurs later in the text, automatically translates it the same way. Exactly how the translation memory functions, and how useful it is, depends on the specific software being used, but the general principle remains the same. This feature is particularly useful in business reports and other texts which use frequently use the same phrases over and over again. It can be highly beneficial for speed, because once a particular phrase enters the translation memory, it can be translated over and over again without the translator even needing to type, only to stop and check that the use of the translation memory is appropriate in that instance.
Translation memory is not infallible, however. In texts with a lot of numbers and statistics, it can be a positive menace, because, for example, it has a tendency to remember that you translated “25%” as “25%” and then want to translate “20%” as “25%”. Take extreme care when translating numerals with a translation memory! In the case of Wordfast Anywhere, this can be solved by pressing the “copy source” button whenever you encounter a purely numeric segment – but when numbers and text are mixed, you still need to be careful. Also, translation memory often uses the same punctuation and capitalisation as the first instance even if it’s different in the current one. Sometimes, a translation memory can be more of a hindrance than a help, but they do tend to get better the more input you give them.
Machine translation is the computerised translation which the CAT software uses to translate the majority of segments for which translation memory is unavailable. Using machine translation, the CAT software produces an equivalent segment in the destination language for every segment in the text. Thus, the translator’s role is often more like that of a proof-reader, checking the computer’s work, which obviously is much faster than translating everything from scratch.
Machine translation is notoriously fallible, however – as anybody who’s tried to translate something using Google Translate and ended up with a page of gobbledygook can attest! In some cases it’s simply a case of an alteration of word order or a slight correction of grammar being necessary, in others the machine translation produces total nonsense and the whole segment will have to be translated from scratch using human translation skills. More dangerous is when the software produces a seemingly perfect translation which in fact is not an accurate communication of the original meaning. This is easy to just skip over, so the double vigilance of checking the segment against the original and then proofreading the whole translation alongside the original is essential to avoid allowing such errors to slip past.
CAT software is indisputably a powerful ally for almost any translator. Although, as we have seen, it needs to be used with care and attention, and it certainly won’t translate the text by itself, it is an invaluable aid in comes to producing professional, high quality translations; especially when working to tight deadlines. Humans are far from obsolete, though – without a skilled translator, even the most expensive CAT software will not produce significantly better results than a free online machine translation.