Linguistic QA (LQA)
A structured in-context review that catches translations which are inaccurate, awkward, inconsistent, too formal, or wrong for the screen.
What it is
Linguistic QA, usually shortened to LQA, is a structured review of translated content for accuracy, grammar, fluency, terminology, tone, and fit with the screen around it. A reviewer records each problem by string, category, and severity. MQM, the Multidimensional Quality Metrics framework that succeeded the older LISA QA Model, gives teams a shared error vocabulary instead of one vague quality score.
Run LQA after translation is visible in a realistic build, with the source, translation context, style guide, and termbase beside it. Use native or near-native reviewers who know the product and target market. Fix critical meaning errors first, then feed approved corrections back to the translation memory and terminology database.
Gotcha: LQA is not the same as localization testing. A linguist can confirm that a sentence is excellent while a functional tester catches the clipped button or broken locale switch. Also separate defects from personal preference: reviewers need severity rules and an arbitration path, or every pass merely rewrites the previous translator.
Ask AI for it
Run an MQM-based linguistic QA review on this localized build. For every issue, output the string ID, source text, target text, screenshot or route, MQM category, severity, explanation, and corrected target text. Check meaning, omissions, grammar, fluency, register, punctuation, placeholder integrity, and compliance with the supplied TBX termbase and locale style guide. Do not mark stylistic preference as an error unless it violates one of those references. Summarize critical, major, and minor counts separately.