27 Nov The human touch in translation
The web and the tools we use are getting more sophisticated by the day. Yet one area of our digital lives that has been impervious to advances is translation. Yes, there is a lot of talk of the advantages of using a machine over human translation—faster, more affordable, more language choices, etc. Yet who has not come away from a Google translation knowing that they were trying to say got lost in translation?
Machine translations are good for simple substitutions where there are few words or for formulaic translations, such as government and legal documents, patents, weather reports, and other documents from professions where the language is standardized. However, when translating other texts, things like idioms, regionalisms and colloquialisms tend to get in the way of a machine’s ability to pull out the right translation. Just try to translate “bangers and mash” (sausage and mashed potatoes in the United Kingdom) or “pop” (fizzy soda in some parts of the United States) in Google translate and see what comes out. Even “fizzy soda” comes out in French with something completely incomprehensible.
Add to this the fact that you can propose a “better” translation in Google, and you are opening a whole other can of worms. Anyone can put anything in as a “better” translation and opens the door to practical jokers who put anything and everything, but clearly not what it should be.
Yet translation software (beyond that of online translators such as Google translate, Reverso, etc.) does have its advantages, especially where long and/or standardized texts are concerned. Such software ensures that the target language and terms are also standardized and leads to fewer errors in long translations. That said, any translation done through any machine, be it online or software, still requires a human touch to ensure that it is error-free. Human translators do not just translate, they also interpret what they are reading, edit their translation to ensure style and voice consistency, and proofread the final text to remove any lingering errors—all tasks that cannot (perhaps yet) be done exclusively by machines.