Translation: Human vs. Machine
It’s easy to understand why people might use machine translations instead of human-generated ones. For one thing, a lot of translation software is free and easily accessible online. But machine translation has its pitfalls. For example, in a recent article, author and translator Nataly Kelly points out that computers don’t understand context, which can drastically change the meaning of a word or phrase. She adds that translation isn’t just about finding equivalent words, but capturing meaning and, in some cases, writing style – something machines can’t do.
As a writer and translator, myself, I agree. It’s not that I don’t sometimes use technology (online bilingual dictionaries can be a godsend). But I know that, ultimately, it’s my own knowledge and, at times, creativity, that will make a good translation.
If you haven’t had extensive experience with machine translation, you may not believe me. So I decided to do an experiment. I chose five different kinds of phrases in French, and wrote my (or in one case, another human’s) translation. Then, I typed the French phrase into two of the most popular free online translation programs (for various reasons, I won’t name names). Here’s what happened:
1. J’ai faim.
Human translation: I’m hungry.
Machine translations: I’m hungry./I am hungry.
This one didn’t surprise me too much; it’s a pretty simple statement, although I wondered whether the verb avoir (“to have”) would trip up the machines. They came through swimmingly.
2. Aimer, ce n’est pas se regarder l’un l’autre, c’est regarder ensemble dans la même direction.
Human translation: Love does not consist in gazing at each other but in looking outward together in the same direction.
Machine translations: Love, is not looking at each other, it’s look together in the same direction./ Love does not consist in gazing at each other, it is looking together in the same direction.
I chose Lewis Galantière to represent Team Human for this translation of Antoine de Saint-Exupéry’s famous quotation, since his version is the one you’ll find pretty much everywhere. The machines surprised me by their varied responses. I have to admit that Program 2 did a pretty good job, but Program 1’s version is littered with grammar mistakes.
3. Ramène ta fraise!
Human translation: Hurry up!
Machine translations: Bring your Strawberry!/Back your strawberry!
Although fraise does literally mean “strawberry”, here it’s a question of context. Idiomatic expressions are one of the key reasons why it’s important to have a human translator.
4. Je le kiffe grave!
Human translation: I’m, like, totally in love with him!/I’m soooo into him!
Machine translations: I kiffe grave!/I kiffe serious!
This contemporary slang phrase is somewhat hard to translate, even for a human like me who watches a lot of French reality television, so I didn’t expect much from the machines. But I was disappointed that, despite the verb kiffer being pretty ubiquitous in pop culture, it wasn’t recognized by either program.
5. Comme dirait Dracula, j’irais bien boire un cou.
Human translation: As Dracula would say, I could go for a neck. -> Let’s make like a Ghostbuster and go get some spirits.
Machine translations: Looks like Dracula, I’d go drink a neck./As would say Dracula, I’d well drink a neck.
Strangely, the machines were totally lost even when making a literal translation. But the biggest challenge here is that the sentence is a pun: the word cou (“neck”) sounds like coup, which in the expression j’irai bien boire un coup means “drink,” as in “I could go for a drink.” As Kelly suggests, a translator’s job would be to take this phrase and make it just as funny (or groan-worthy, depending on your feelings about puns) for an Anglophone audience. In some cases, this could mean having to replace the pun with another one entirely, as I did…hopefully not too badly.
Winner: Humans. …Although, being unable to generate a terrible pun, the machines win in the dignity department.