Translation Quality Review

Translation quality review

A translation can be accurate and still wrong: too formal, too literal, or using a term the industry does not use. Review needs a rubric, because the common feedback of it sounds odd is unactionable.

Method

1. Check accuracy against the source meaning, not the words. A literal rendering that loses intent is a failure, and a free one that preserves intent is usually right.
2. Check register and formality. Languages with formal and informal address force a choice, and inconsistency within one product is more jarring than either choice would be.
3. Check terminology against the glossary. Product terms and industry terms must be consistent across screens, which is where multiple translators diverge (see translation-workflow).
4. Review in context, on screen. Truncation, awkward line breaks, and a string that reads differently beside its heading are only visible in place.
5. Verify placeholders and plurals survived. A dropped placeholder or a missing plural form is a defect that renders as broken text or crashes (see plural-and-gender-rules).
6. Have a native speaker who knows the domain do the final pass. Fluency alone misses domain conventions, and domain knowledge alone misses tone.

Boundaries

• Review judges the translation; it cannot repair source copy that was unclear to begin with, and ambiguous English produces ambiguous everything.
• Reviewer disagreement on style is common and needs a style guide to settle rather than repeated debate.
• Automated quality scores catch mechanical errors and cannot assess register or cultural fit.