Thread by @ceddytrad

I say it often: word-for-word translations are often stilted, unnatural, and ripe for improvement.

An example from a recent #translation:

Source: “Le niveau de gravité selon la matrice ci-dessous permet de définir l’ordre de priorité pour la résolution des demandes.”

1/5

#xl8

My first draft was passable:

“The severity levels set out in the following table render it possible to establish the order of priority for handling support requests.”

Not awful, but there is clearly room for improvement.

2/5

Just for laughs, I ran it through DeepL:

“The severity level according to the matrix below is used to define the order of priority for the resolution of requests.”

(Wow! What a horrible, awful, no-good word-for-word #translation! What does that even mean?)

3/5

Later, I cleaned up my translation to make it shine.

More clear and natural, it says the same thing without sticking too closely to the source:

“Support requests are prioritized in accordance with their severity level as described in the following table.”

4/5

The takeaway?

🔷Even a decent first draft often just gets the gist of the original. Proofreading is key.
🔷NMT is not nearly as good as anyone says. In fact, often it doesn't even give a good starting point.
🔷Translate what is meant or intended, not what is said literally.

5/5