Holidays at last! You’ve checked TripAdvisor for reviews and selected a nice quiet hotel. But have you ever wondered why the comments are often so badly written? It’s actually not a wonder at all. TripAdvisor says, “This review has been machine translated from English,” which is another way of saying that zero effort has been put into the quality of the translation. Incidentally, the cost of the translation is also zero. Is this the end of the translation industry? Can all translators be sent on a permanent holiday now?
Syntax error in transcreation
No is the answer to both questions. Although language technologies are constantly improving, machines are not yet capable of thinking creatively. For important headlines such as our slogan “We’ll put in a good word for you,” the machine-translated suggestions by Google are useless: “Nous avons mis dans un bon mot pour vous.” Mon Dieu! It’s bad enough to make your French teacher’s hair stand on end. In many cases, it’s worth rewriting the text in the foreign language, rather than translating it. “Ne cherchez plus vos mots, ils sont chez nous,” for example, is obviously better. Having said that, translation software has become indispensable for translators nowadays – and not just to save costs.
CAT tools increase consistency
Companies and translation agencies are increasingly deploying CAT tools, which stands for Computer Aided Translation tools – programs in technical jargon – and include translation software such as SDL Trados, Déjà Vu, Star Transit, Wordfast and Across. The most common software is SDL Trados Studio; the Microsoft Office of the translation industry. Which is why Supertext is an associate partner of SDL.
Essentially, all translation software functions in the same way. The program reads a Word document or another file format and divides the text into separate sentences or paragraphs, called segments. The translator then translates each segment directly in the program and the translations are saved in a database, called the translation memory. If an identical phrase comes up in subsequent texts, the program suggests the saved translation. For similar phrases, a concordance search can be carried out on all segments saved. This accelerates research and promotes consistency.
In addition to the translation memory, a terminology database can also be created, which stores and maintains accurate translations of specialized terms with comments. A blacklist of terms to avoid can also be maintained here.
What’s the catch?
This may all sound great, but there are also disadvantages. In fact, there are many. By separating the text into segments, the translator is trapped in the structure of the source text. This can be mitigated by inserting the translation into the segments without regard to the corresponding source text, but then the quality of the translation memory will suffer. Furthermore, the translator does not translate in the layout – the text is displayed in an Excel-like table, segment by segment. The overview is lost, which can result in a translation that doesn’t flow well as a whole text. So the software can be an efficient tool for technical texts with many repetitions, but is only useful for consistency and terminology when it comes to marketing copy. Because the ultimate sign of a good translation is when the reader doesn’t realize it’s a translation.
Going back to the reviews on TripAdvisor, the trend is to automate communication when the purpose is purely for comprehension. And machine translations are being continuously updated with translation memories and corrections – carried out by human beings, of course. Try translating Facebook status updates from friends via Bing and have a good laugh.
Cover image via Flickr: DSC 0521 – Blake Patterson (CC BY 2.0)