Grade a translation by marking each fault's span, kind and weight instead of your impression.
This work is graded inside a frame the listing rarely names. MQM sorts every fault into seven kinds and weighs each one on four levels. The kinds run from accuracy and terminology through wording, style, local custom, audience fit and layout. The four weights — neutral, minor, major, critical — are counted as 0, 1, 5 and 25. Three validators often mark one sample without seeing each other, so your marks have to stand on their own.
- Mark the span first, then name the kind of fault, then say how bad it is.
- Kind and weight are two decisions, not one: a spelling slip is not a dropped warning.
- Critical means a reader would act on it wrongly, however small the words are.
TaskAn English health notice and a Hindi version of it, with a written reason due beside the choice.
Weak approach The second version reads better and sounds like a doctor really wrote it, so it wins on wording. The reason box says exactly that.
Stronger approach Three marks, each on its own words. 'खाली पेट' is gone: left out, an accuracy fault, and critical, because a reader takes the medicine wrongly. 'डॉक्टर को दिखाएँ' where the notice said to telephone is a wrong meaning, major. 'आप' where the notice used 'तुम' is a style fault, minor. Only then are the two versions compared, fault against fault.
Why the stronger approach works Both reviewers preferred the same version. Only one of them left three lines a buyer can count, argue with, and add to a score.
Try a changed situation The same work, but the text is now an app's Hindi menu checked against the English one. Where does a mark go?
Make something yourself Take one short English passage and a Hindi translation of it. Mark every fault with its span, its kind and its weight. Write the justification out of those marks, not out of your impression.