Invest in Transfluent - our public share offering is open until the end of May, 2017! ×

4 Things Computer-Generated Translations (Still) Can’t Do

With recent developments in translation software and technology, such as Google’s newly released real-time speech-to-text translator, it seems more possible than ever that computers will entirely replaced human translators. Indeed, automatic translation services have made great strides since their humble beginnings. However, despite the recent buzz surrounding flashy new online translators, there are several areas in which computer-based translators still don’t — and most likely never will — make the grade. Automatically generated translations are useful for the basics, but not much beyond that: if you’re looking for a real, high-quality translation, it’s crucial that it be done by a human and not a computer.


This could be you if you use an online translator instead of the real deal. Image via CollegeDegrees360 / flickr

  1. Computers don’t understand context

In the early days of online translators, translations were done word-by-word, resulting in humorously incoherent sentences like the following, taken from an Argentine tourism website:

“In the province of Buenos Aires they exist around 150 helmets of stay with lodging capacity. Of modest to luxurious, all offer varied alternative for all the tastes and budgets. It is possible to be enjoyed an only day of field or one more estadía prolonged.”

While I applaud the fact that “Buenos Aires” was not translated as “Good Airs”, it’s clear that this translation is only a small step above indecipherable drivel. This is because word-by-word translations don’t take context into account. For example, in the above paragraph, the word casco was translated literally as “helmet”, its most common meaning. However, it’s clear that the intended referent was a casco de estancia, which translates more accurately to “ranch” or “estate”.

Nowadays, automated translations have strayed from the unreliable word-by-word method used in the past. Indeed, most consider the context of the surrounding words to deduce a particular words’ meaning. For example, a decent translation software will know that running means different things when we say “running water” and “running away”.

Still, despite these advances, computer still have no way to consider the broad, overall context of a piece of writing. Who is the audience? What is the author’s attitude towards the subject? How can the author’s tone be categorized? These questions are among the first and most important ones considered by professional translators, as they can drastically affect the overall flow and style of the piece. But they are lost with automatic translations, which simply cannot take such questions into account.

  1. Computers don’t have a sense of humor


This computer can do a lot of things. Understand humor is not one of them. Image via Kristoferb / Wikipedia

Humor is a uniquely human and highly subjective phenomenon. It depends on timing, tone, and carefully chosen words, and as a result, computer-based translators are notoriously bad at it. If you’re translating an academic paper or instructions for using a microwave, this might not be a huge problem. But if you’re trying to translate a casual blog post or content to share on social media, automatic translators’ inability to understand humor could be a serious hindrance to your effective communication.

As an example, consider the following sentence: Online translations are so awesome at conveying sarcasm; they totally get it! While you, a human being, easily pick up on the fact that the utterance is dripping in sarcasm, Google Translate won’t, and as a result, your translation will suffer. And if your content involves puns or wordplay, the literal translations provided by automated systems will almost certainly not make sense in your target language. Indeed, successfully conveying humor requires a basic understanding of what humor is — something possessed only by real-life humans.

  1. Computers aren’t creative

Language is not always literal. Often times, we use clever language or idiomatic expressions for which a literal translation will not suffice. In fact, frequently this will involve using a completely different word — or even sentence — to convey the same sentiment. For example, in the French-language book I Hate Hockey by Francois Barcelo, a father buys his son, named Jonathan, a bicycle that is yellow (jaune), so that he can write “À Jaunathan” on his birthday card as a joke.

When translating this book into English, (human) translator Peter McCambridge wanted to conserve the pun, but found it difficult, as “yellow” and “Jonathan” hardly sound the same. To fix the problem, he changed the color of the bike to white, so that the birthday card read, “All white, Jonathan?” Had the same passage been translated by a computer, the bicycle would remain yellow, and English-speaking readers would be utterly confused as to why Jonathan’s father thought that it was funny to do so.

  1. Computers lack cultural awareness

The popular HBO series In Treatment (which itself is a translated version of an Israeli series) follows the therapy sessions of several individuals with various life problems that need sorting out. A Spanish-language version called En Terapia was produced in Argentina a few years later. It featured the exact same characters and plots, which were modified minimally to be culturally appropriate and realistic.

One of the plot lines follows Laura (Marina in the Argentine version), who is in love with her therapist, as we learn in the first episode. Interestingly, in the Argentine version, Marina flat-out says, “Estoy enamorada de vos” (“I’m in love with you”), whereas in the English-language version, the grand reveal is done subtly, through a series of circumlocutions and shifty facial expressions.


Marina (left) and Laura (right) from the Argentine and United States versions of En Terapia/In Treatment. Images via HBO and TV Pública

This was a conscious decision on part of the translators who wrote the Argentine script. Whereas those living in the United States are often reserved or implicit when dealing with their feelings, there’s very little of this beating-around-the-bush in Argentina. Thus, the translators decided that it would be more culturally appropriate for Marina to state her feelings directly instead of giving hints. A computer-based translator — even one programmed to use the Argentine dialect — cannot serve as a cultural liaison in the way that humans can; it could never make such a judgment call.

So what are online translators good for?

So — online translations can’t take context into account, they don’t get sarcasm, they lack creativity, and they don’t understand cultural awareness. But they’re not completely useless. Indeed, they can be useful for finding out the basic meaning of something when producing a high-quality translation isn’t important. And Google Translate’s app, which allows you to take pictures of street signs and translate them in real time, can be a life-saver for tourists.

But if you really want a high-quality translation that conveys the true essence of the source text and uses carefully chosen, culturally appropriate language in the target text, you’ll need a real-life professional human translator to do so. Yes, computerized translations have come a long way — but they’re still far from replacing the human element, which remains the most crucial component in ensuring a successful translation.


Paul writes on behalf of Language Trainers, a language tutoring service offering personalized classes on an individual or group basis. You can check out their free foreign-language level tests and other language-learning resources on their website. Visit their Facebook page or contact if you have any questions or if you’d like more information.