Translation is a complicated art. Anyone who’s studied a foreign language or taken a handful of linguistics courses knows this: there is no one-to-one ratio of words and concepts, and cognates are not always perfect matches. The fact that English loves to borrow words and phrases from other languages (e.g., c’est la vie, wanderlust, karaoke) should be proof enough that languages don’t match up perfectly. So translation, particularly fiction translation, is a complicated undertaking.
For this reason, translation should not be done by only one person or by AI. As Tim Parks discussed in the New York Review of Books, translation is not always done well. If you speak the original language a translated book is printed in, you may find continuity errors brought about by issues with direct translation. For this reason, translation cannot be done by only one translator. Different translators will have different backgrounds with the source language and different ideas about how things should be conveyed in the target language.
If you want to avoid bias, you might wonder, What if AI did it? No human bias there. But with translation software, there’s a different kind of bias that appears: statistical bias. Translation software is complicated, and the field has grown quite a bit since it first began. Statistics isn’t the be-all and end-all of translation, since language is generative and more complicated than algorithms that calculate statistical likelihood. However, after the structural analysis has occurred, statistics does play a role. Translation software, at this point, uses the most likely cognate for a particular word, or the most likely equivalent for a short phrase. This is great for when you’re in a pinch and you need to learn how to ask where the bathroom is. It doesn’t do so well with ambiguity, figurative language, or slang, all of which may be found in a piece of fiction. Perhaps one day translation software will reach a point that allows for nuance, but for now, translation should be a collaborative effort in order for it to be done well.
Perhaps this is why so few translated novels are published in the US each year. Translation is complicated, expensive, and risky to publishers. Some have even said that Americans aren’t interested in reading translated works—for one thing, there are plenty of books being published locally, and for another, books from other countries may feel too alienating.
Additionally, out of the few translated works published annually, 40 percent come from just a handful of Eurpoean countries. There are thousands of authors’ voices from all around the world that American audiences are missing out on. On the bright side, small presses are contributing to a growing market for translated works. So if you feel like broadening your horizons and trying something new, go out and find a translated book to support small publishers.