FAQs - AI Quality Evaluation

AI Quality Evaluation is an AI-powered tool designed to assess the quality of translations and other content. It combines AI efficiency with human expertise to provide accurate and fast evaluations, helping businesses ensure high-quality content and reduce costs.

The primary purpose of AI Quality Evaluation is to act as a quality checker for translations. It evaluates uploaded content after it has been translated, helping businesses determine if the content meets quality standards. The tool provides a score based on the quality and suggests whether further human review is necessary.

After uploading a document, the content undergoes translation through phases like Leverage TM (Translation Memory) or Machine Translate, followed by AI Quality Evaluation. The AI system reviews the translation and provides a quality score, categorising segments into different levels (Best, Good, Acceptable, Bad, No Score, Untranslated).

AI Quality Evaluation supports a variety of file types. Please refer this for the supported file types, https://lingotek.atlassian.net/wiki/x/BoCgwg.

To use AI Quality Evaluation, you need a project workflow template that includes this as a system phase. You can create a new template or add the AI evaluation step to an existing workflow. It is recommended to place AI Quality Evaluation after Translation Memory (TM) or Machine Translate phases.

You can add the AI Quality Evaluation phase as a custom phase to any existing workflow template by editing the template under Templates > Workflow Templates. Place it after Leverage TM or Machine Translate phases to ensure translations are processed before evaluation. Make sure the workflow is saved and applied when creating new projects.

The default threshold is 100%, meaning only perfect scores are approved automatically. You can adjust this threshold per project under Project Workflow > AI Quality Evaluation to balance automation and manual review workload. Lowering the threshold will approve more segments automatically but may reduce overall quality assurance strictness.

After creating a project using a template that includes AI Quality Evaluation, you can upload the document to the system. Please ensure the document meets the supported file type requirements.

AI tokens are consumed during the quality evaluation process. The number of tokens required depends on the content’s complexity and the length of the document. The token cost is visible in the document summary, and you will need to purchase tokens to initiate the evaluation.

AI tokens are currently priced at $0.02/token (USD).

Once the content is uploaded and translated, the system will calculate the token cost for the evaluation. You can purchase the required tokens by clicking the "Purchase AI Quality Evaluation" option, and the evaluation process will start once the purchase is complete.

Once purchased, the system will process the quality evaluation, and you will receive a quality score represented as a percentage. The score will reflect the overall quality of the translation, and segments will be categorised (Best, Good, Acceptable, Bad, No Score, Untranslated) to give detailed feedback.

Yes, you can configure additional human review phases after the AI Quality Evaluation if necessary, based on the evaluation results. This flexibility allows you to adapt the review process based on the AI-generated quality score.

After evaluation, segments are categorised as Best, Good, Acceptable, Bad, No Score, or Untranslated. These categories help reviewers focus their attention efficiently—Bad or Untranslated segments usually require human intervention, while Best or Good can often be approved with minimal checks.

No, segments are locked and cannot be edited in the AI Quality Evaluation phase. To modify translations, use subsequent review phases. Segments meeting or exceeding the quality threshold are locked in the Workbench with an “AIQE Approved” indicator, simplifying manual review.

After making edits, the previously calculated AIQE scores become invalid. To re-evaluate, select the affected targets under Targets > Actions > Purchase AIQE to trigger a fresh AI Quality Evaluation. This ensures scores reflect the latest translation content and helps maintain quality assurance integrity.

It’s best practice to run Leverage TM and/or Machine Translate phases before AIQE to ensure translated content is finalized. Position AIQE after these phases in your workflow template to accurately assess machine-generated or leveraged translations. This order prevents unnecessary AIQE token costs on untranslated or draft content.

AI Quality Evaluation supports 156 languages on the Enterprise platform, ensuring a broad range of language pairs for businesses working on multilingual content. Please refer Supported Languages .

In such cases, the system skips AIQE automatically. Instead of token cost, you will see messages like “Language not supported by AI Quality Evaluation” or “AI Quality Evaluation not needed for exact match” in the Resources column or purchase dialogs. No AI tokens will be consumed, and the evaluation phase is effectively bypassed.

AI Quality Evaluation is designed specifically for translations and content that requires accuracy in language. It works best when integrated into translation workflows, but it can also be used to evaluate other types of content as per the platform's capabilities

You can access AI Quality Evaluation through the Enterprise platform’s Workflow Templates or by adding it as a custom phase to an existing project. Ensure you have the necessary subscription or plan that supports this feature.

Yes, after completing the evaluation, the results will be displayed with a quality score and a detailed breakdown of how each segment was evaluated. You can view the categorization of each segment (Best, Good, Acceptable, Bad, No Score, Untranslated) by clicking the score in the Resources column.