Available for Custom Workflows, the Quality Evaluation step uses AI to assess your translations and assign a score based on their quality. You can use the evaluation results to choose how the workflow continues.

When used as the first step in a workflow, it can run on the following triggers:

• New terms
• Reference language updated
• Scheduled
• Manual

If a Default Reference Language is set on the project, it is automatically used for the evaluation. You can also choose the Project Terms, the language inherited from the previous step, or any other language in the project. If your terms are keys or string IDs, a project language is a better choice for the reference.

For the evaluation, you can choose:

• the languages inherited from the previous step
• a custom selection of languages
• all the project languages

You can then set what happens based on the evaluation results:

• never stop the workflow
• stop the workflow if QE errors found
• stop the workflow if no QE errors found

Translations with a score below 100 are sent to the Failed batch, along with the reasons given for the score. Translations with a score of 100 are sent to the Passed batch. The steps you add to each batch only receive the translations assigned to that batch.

The step skips translations in the reference language, translations that have already been evaluated, proofread translations, empty translations, and non-translatable terms. When the step stops the workflow, all following steps are skipped.