“When I am in trouble, Maria? Maria comes to me and sings words of wisdom: let’s be it, let’s be it”
Those are the lyrics to The Beatles’ timeless hit “Let It Be” run through Google Translate from English to Japanese. Then back into English. It might sound familiar, but you have to admit that something’s literally lost in translation.
These days, most of us have tried to use a machine language translation like Google Translate–and we’ve seen the limitations.
That said, there’s no denying that these programs have improved drastically over the last several years. It’s lead many people to wonder if a day might come when machine translation could surpass human translation.
The answer has more parts than you might think.
But first, we have to answer one preliminary question…
What Makes Good Translation?
At its face, translating between one language might sound like a pretty simple idea. In every language, words have definitions, right? All you have to do is find the words with the same definition and switch between them.
If this is all you want, AI language translators might be able to make it happen. But if you want to communicate clearly between languages, that won’t get you very far.
There are two theories of how to perform translation: metaphase and paraphrase.
Metaphrase is a literal word-for-word translation. Paraphrase, on the other hand, is an idea-for-idea translation. In an idea-for-idea translation, you try to find the closest equivalent idea between the two languages.
Most translators agree that good translation is a mix of the two, and that’s bad news for automated language translators for a number of reasons.
Context Is Everything
Language is a complex thing. To assume that you can preserve meaning between languages with a simple word-to-word translation can miss a lot of this complexity.
For example, let’s take the word “run.” It’s a pretty simple word, right?
Except that it has hundreds of different meanings.
Someone could be moving at a brisk pace. Or, they could be letting water through a tap. Or there could be a tear in their stockings.
To properly translate a word, you need to have an understanding of the context surrounding that word. Without the context, it’s impossible to know whether “bat” means a flying mammal, a club used in sports, or to jab at something.
While AI translators have gotten much better at identifying context, they still have a long way to go before they match human translators, let alone surpass them.
Culture Matters–A Lot
Language doesn’t exist inside of a vacuum. Every language is strongly influenced by the culture the language was born from.
If you don’t understand a language’s culture, you run the risk of making a translation that is terribly offensive–or at least very confusing.
Our languages are filled with slang and idioms that make no literal sense. If you were to try to tell a non-native English speaker to hold their horses, there’s a good chance they’ll look at you like your head is on backwards.
Similarly, you’d probably be confused if a French person told you that they had the cockroach. But that just means that they were feeling depressed.
These cultural details inform much of our language. And as advanced as Google Translate might become, programming an AI translator with every bit of cultural knowledge in the world just isn’t feasible.
If you consider how quickly language changes, that programming would be a constant struggle.
Machine Language Translation Doesn’t Understand Nuance
When you’re trying to define words, there are two principles at play: denotation and connotation. Denotation refers to the strict dictionary definition of a word. Connotation refers to the social and emotional implications attached to those words.
Looking at the denotation of the words “cheap” and “inexpensive” basically equates them to the same thing. But the connotations are very different. We understand that the words feel different, even if we can’t explain why.
Machine translators don’t really do well with the undefinable feelings attached to words, which makes translation difficult.
And when you’re translating things like technical manuals for manufacturing equipment, you can’t afford to trust that to a machine.
Localization Is Just As Important As Translation
One of the trickier parts of translation is localization. A single language might have hundreds or thousands of local dialects, each with their own unique idioms and phrases.
An American might feel like they can visit New Zealand without having much trouble communicating. But after a few weeks of wondering what a box of fluffies is, you’ll realize how wrong you were.
Sometimes, even speakers in the same country speak wildly different dialects. Imagine a conversation between a person from Brooklyn and someone from the bayou outside of New Orleans.
Just because a translation works for one dialect doesn’t mean speakers of another dialect will have an easy time understanding it.
But localizing a translation between dialects often takes a human touch. In fact, a good sense of localization is what makes the difference between a good translator and a great translator.
Man vs. Machine
Machine language translation can be a useful tool. Apps like Google Translate can come in handy if you’re traveling abroad or trying to talk with a neighbor from another country.
But if you need a quality translation that captures the true meaning and nuance of the original language, there’s no substitute for a human translator.
If you have a project that needs a skilled translator, contact us for a quote.