Read for meaning, not flow

Module 1: What AI already does in translation

Course outline

Quick check

Guess first, no penalty. A machine translation reads smoothly, with perfect grammar. What's the biggest risk?

MT (machine translation) tools are built to produce natural sentences. That's why their errors survive a quick read: a wrong sentence that reads well doesn't look wrong.

So read for meaning, against the source, line by line. Watch hardest for negations ("not", "never"), numbers, units and conditions ("only if", "unless").

Spot the changes

The English source of a patient leaflet: "Take one tablet in the morning. Do not take this medicine more than twice a day. Keep out of the reach of children. Stop taking it and call your doctor if you get a rash."

Spot the mistakes

Tap the two Spanish lines that don't match the source. Tap a line that's fine and you get a hint.

MT draft (Spanish)

0 of 2 found

A check prompt for the risky words

A prompt is the message you type into the tool.

Here is a source text and its translation: [paste both].
List every negation, number, unit and condition in the source.
For each one, quote the matching words in the translation, or write "missing".
Don't rewrite anything.

Fill the gaps

Fill the gaps so the check catches a reversed meaning.

Prompt

List every in the source. For each one, . .

Try it

Your steps, about 2 minutes

0 of 1
  1. 1.Run the check prompt on one translated text from your work, and fix anything marked "missing". Add the prompt to your post-editing kit.2 min

Practice

Question 1 of 2

You review a machine-translated paragraph that reads smoothly and has no numbers. Which check is most likely to catch a sentence that reverses the source?

AI tutor

Ask about this lesson and get coached through the Try it task. The tutor is part of Learn Pro.

See Learn Pro