AI Features

AI Features

Guide to Using AI Quality Boost and Quality Evaluation

Overview

AI Quality Boost is a suite of specialized AI agents that automatically edits translations to ensure they meet your specific terminology and style requirements, all while preserving the original meaning. This feature reduces the need for extensive human review by automatically improving the quality of AI translations. AI Quality Evaluation understands the quality of your translations with a quality score. The tool evaluates content for fluency, accuracy, and consistency, helping you decide if it needs further review.

Purpose

  • AI Quality Boost is designed to proactively enhance translation quality by automatically editing and refining machine translations. Its primary function is to align content with your specific terminology and style, preserving the original meaning and significantly reducing the post-editing workload for human reviewers.

  • AI Quality Evaluation acts as an essential quality checker within your translation projects. By scoring your content, it allows you to make informed decisions about whether to publish the translation or send it for further human review, thereby reducing manual effort and saving time. The process consumes AI tokens, which are automatically estimated during the workflow.

Key Features

  • Integrated Workflow Step: Both AI Quality Boost and AI Quality Evaluation are seamlessly integrated into your project workflow as dedicated phases.

  • Intelligent Processing: The system is designed for efficiency. It automatically skips segments that have 100% matches from your Translation Memory (TM) and only processes segments that were generated by machine translation or require changes.

  • User Progress Tracking: You can monitor the progress of the AI Quality features on the Process Queue Page or the Processes popover, similar to how you track other translation operations.

  • Quality Score Threshold: In the Workbench, you can set a quality score threshold. Translations that meet or exceed this score are automatically locked and considered approved ("AIQE Approved"), minimizing manual review.

  • Report & History: After completion, you can view a simple summary report, including a pie chart of the results. Historical information is also available in the segment history view.

  • Resilient Workflow (Skip Logic): If the AI service is unavailable or unusable for your job (e.g., language not supported or a platform issue), the system is designed to gracefully skip the phase and allow your project to proceed to the next step of the workflow without interruption.


Getting Started

How to Enable AI Feature in Enterprise TMS 

  1. Go to System then select Communities

  2. Search for the Community name and click on it to open its settings.

  3. Go to the AI Feature section 

  4. Click Edit Key, enter your verify API Key Click Save. 

  5. Tick the corresponding box to enable the AI Features for the selected community.

Creating Project Workflow Templates with AI Feature
This AI Feature is a system phase that must be included in your workflow template.

  1. Go to Templates then Workflow Templates to create or edit a workflow template.

  2. Add AI Quality Boost and AI Quality Evaluation as phases.

  3. Place AI Quality Boost after your Leverage TM and/or Machine Translate phases and before the AI Quality Evaluation phase.

 

Create and Upload the Document

  1. Create a new project using the workflow template that includes the AI Feature phases.

  2. Configure your initial settings for the Leverage TM and Machine Translate phases.

  3. In the Project view, under Project Workflow and AI Quality Evaluation, adjust the quality threshold if necessary. Translations scoring at or above this value are automatically approved.

  4. Upload your document (must meet supported file type requirements) or input the content.

  5. Select the target language(s) supported by the AI Feature.

 

AI Quality Boost and Quality Evaluation Process

  1. The content first passes through the Leverage TM and/or Machine Translate phases.

  2. The translated content is then sent to AI Quality Boost for token estimation. The system will display the estimated token cost for the request.

  3. Approve the Token Cost: To proceed, go to Action then select AI Quality Boost Cost View. You can approve or reject the boost for each target language using the dropdown list.

  4. Once finalized, click the Submit Decision button. The overall token cost will be deducted from your available balance.

  5. After AI Quality Boost completes, the Resources message will show the percentage of changes made by the AI. You can click this text to view the AI Quality Boost Report.

  6. Following the Quality boost, AI Quality Evaluation will estimate its own token cost and require your approval in the Action then select AI Quality Evaluation Cost view.

  7. Once phases are complete, click the text under Resources to view the final Quality Evaluation Report.

     

Note:

  • If a language is not supported or if the document content is an exact match from the Translation Memory, the system will skip the phase and no AI tokens will be consumed.

  • Token balance can also be viewed by navigating to the "Community" tab and clicking on "Customization." then select the "AI Feature".

 Review Translations with AI Feature results

  1. Once the quality score is ready, you can proceed to the Review phase.

  2. Clicking the Review text redirects you to the Workbench.

  3. Approved Segments: If a segment processed by AI Quality Boost meets or exceeds the AI Quality threshold you set, it will be locked in the Workbench. Hovering over the lock icon will display the quality score and the "AIQE Approved" message.

  4. Segments Needing Review: If a translation score is below the AI Quality threshold, the segment by default is unlocked. The reviewer can then manually update the translation and approve it.


To see these features in action, watch our video demonstration :

AI Features in Enterprise_Help Center.mp4