The agony of machine translations

Let me introduce Robert Falck. A writer and geek himself, Robert has worked with Applingua over the 18 months on a variety of different apps. This week Robert tells us what it’s like to read and work with machine-translated text.

I’m a native Swedish speaker, so I have a somewhat different view on localization and app translation in general. Not being a native English speaker, I’m very tolerant when no translation is available. Instead apps are usually only presented in English and I’m fine with that.

It’s when we start dipping our toes into the realm of machine translated app descriptions that things start getting a little bit frustrating. Seeing a non-English text, mangled by machine translation, as a native of the language in question, can be anything from confusing to downright insulting. You know when something is translated by a real person and you instantly notice when it’s a machine behind the reins. There are just too many nuances and ways to express yourself that nearly only a native or very apt student of a language can pick up, use and understand. This has, so far, sadly not extended to translation algorithms.

When I see an app description, I expect to get an informative and persuasive text about why the app is something I should get. If that text is instead a mish-mash of words normally not normally used and a grammar that makes you question the sanity of whoever wrote it, rest assured I will hit “go back” and look for another app instead. Price is less relevant too, since I can’t see the effort of making a proper presentation of the app in the first place. It just leaves a bad first impression.

Leaving possible lost sales aside, you might even end up with something far more unpleasant on your hands. If we take a look at the example of the translation error that lead to a racist phrasing in a product description, we see that things can get quite bad if we trust the machine a bit too much. Granted, this is a bit of a special case for several reasons, but it does prove that things can turn out very bad sometimes.

Between closely related languages, those which are very close to each other in the same language family, can, with greater success, be machine translated, but it will still not be perfect. Imagine how well it works to go from Swedish to Japanese with a very common machine translator like Google Translate. The outcome does not make much sense, but it does have an entertainment value for anyone who reads it. Not something you want to have representing your new app on the App Store!

I have worked with several machine translated texts over the years and I have colleagues who have had to suffer them as well. Whenever I’m given a project with a machine translated text, it ends up being more of a mystery-solving exercise, trying to figure out what the original text really means, rather than a proofreading or simple spellchecking. Most of the cases I have ended up asking for the original text instead, opting to do a brand new translation instead of working with the horribly broken translation I was given to “correct”. The cost for it all I’m sure you can guess.

You might think you are saving yourself some money and enabling yourself to branch out into more markets, but in reality you are just scaring people away from your app.


Robert Lo Bue

Rob is CEO of Applingua. With over a decade's experience, he is at the forefront of tech localization.

Comments

  1. This like to PRC phone production which have so funny translation. I saw translation from English to russian and for example enstead ‘Save phonebook’ they are translated as ‘Resque phonebook’)))