Glossary quality assessments explained
Applies to: pagenorma S, pagenorma QS
Besides listing terms, the glossary can assess whether each translated term looks right in context.
The four statuses
| Status | What it means |
|---|---|
| Red flag | The translation looks wrong. Look at it again. |
| Needs improvement | It may work but could be better in this context. |
| Appears correct | No problem was found. |
| Not assessed | The pair was not judged. See below. |
Click a status badge to Show assessment details.
The details
- Assessment — the status with a score out of 100. A score of 0 means clearly wrong; 100 means very likely correct.
- Confidence — how sure the AI is of its own assessment, out of 100. It is not a measure of the translation’s quality.
- Issue type — a short label for the kind of problem. Use the issue-type filter to list all terms with the same one.
- Assessed translation — the exact wording that was judged.
- Reason — the explanation.
Why “Not assessed”?
Before judging quality, the AI checks that the source term and the translated term are really a matching pair. It does not assess a pair when:
- one expression is only a cut-off part of the full match (a partial pair),
- both expressions exist but are not meant as translation equivalents (a mismatched pair), or
- it cannot tell from the full sentence.
In those cases the reason explains why the pair is not valid.
What is deliberately not an error
Natural, non-literal translations, and word endings that the sentence requires, are not treated as mistakes. The assessment looks at idioms, figurative language, words with several meanings, cultural and historical terms, inserted content and inappropriate register.
Using it
Start with Red flags: filter by that status, open the reason and context, and decide for yourself. The assessments are a guide for review, not a verdict.
Last updated 20 October 2026