Rumor debunked! Machine translation is actually bad for marketing texts

Have you heard this one? People have been boldly asserting at parties and random get-togethers that neural machine translation is just great for marketing texts.

Well, it’s a nasty rumor.

My theory is that this reflects an inherent bias against non-scientific, liberal-arts subjects that people carry with them from their earliest schooling.

Law, medicine, and the like here are the weightiest of subjects deserving the utmost respect; marketing and other creative vocations are subjects for undecided majors in college.

In reality, marketing texts and creative prose in general are among the most difficult things for neural machine translation to handle.

Why? Simply because marketing texts are highly flexible in their word choice, idiomatic expression, and syntax. They routinely push human expression to the limit. They celebrate language by trying to squeeze the most meaning and nuance out of every word.

In short, DeepL, Amazon, and Google can’t keep up.

One of my favorite slogans is from Lufthansa: “Non-stop you.” DeepL translates this as “Ohne Unterbrechung.” Which is totally absurd.

So maybe that’ll finally put this nasty rumor to bed? (Don’t ask me how DeepL translates this….)

Christopher ReidComment