When using translation quality evaluation, a quality score from 1-100 will be estimated to help you assess the quality of each individual segment that makes up your translated text and is calculated based on the following factors:
- Number of identified issues and their severities: If a translation error is found, it will lower the score of that segment based on its severity. Critical errors have the largest effect on the score, while minor errors will affect it less. Learn more about the severity of translation errors here.
- Issue density: The score will be affected more severely by the number of issues in a segment based on its length. A short segment with more issues will have a more severe impact on the score than a much longer segment with the same number of issues.
Segment-level scoring is available for translation quality evaluation within DeepL's Translation Flow, the translation quality evaluation tool and the early access version of the translation quality evaluation API.
If you are interested in testing this feature, reach out to your Account Executive or Customer Success Manager.
Score breakdown
The score can be broken down into the following assessments:
- 100-95: Excellent
- 94-85: Very good
- 84-70: Good
- Below 50: Poor
Routing
When using the translation quality evaluation within DeepL's Translation Flow and adding a team member as a reviewer, you can set it so that the document is only sent out for review if the translation falls below the quality threshold you have set. For more information on setting up a Translation Flow, see this article.
The translation quality score will only be calculated once per translation. It won't be updated based on any edits.