Matches docked panel
XTM Workbench displays matches in the docked panel. They are ordered by the type of match and then by the date created or changed. The order is as follows:
In-context Exact match (ICE).
Leveraged match 100%.
Repetitions.
TM Fuzzy matches.
Internal Fuzzy matches (in score order and then in date order).
MT matches.
TQI results in XTM Workbench
If your XTM Cloud Administrator has activated the AI Translate & TQI feature for your instance, you will see the TQI results in the Metadata column, in the Matches docked panel.
The translations generated by the AI Translate & TQI feature are marked as Open AI GPT in the Matches docked panel in XTM Workbench, regardless of which LLM(s) – Large Language Models – have actually been used for the feature.
The AI Translate & TQI feature automatically evaluates the quality and accuracy of translations generated by an LLM (Large Language Model). For every eligible segment, the feature compares translations generated by two or three different LLMs and evaluates their similarity. The lower the similarity of these LLM-generated translations, the lower the quality evaluation score (expressed as a percentage).
TQI COLUMN VALUE | DESCRIPTION |
|---|---|
Percentage on a red background and an error icon | The LLM-translated target segment has been evaluated as a Not acceptable (poor-) quality segment. This means that the segment needs correction or post-editing by a Linguist. Hover over the result in the TQI column to see the improvement suggestions. |
Percentage on a yellow background and a warning icon | The LLM-translated target segment has been evaluated as an Acceptable (medium-) quality segment. This means that the segment meets basic standards, but a Linguist might need to make minor improvements to it. Hover over the result in the TQI column to see the improvement suggestions. |
Percentage on a green background and a checkmark icon | The LLM-translated target segment has been evaluated as a Good (high-) quality segment. This means that the segment does not need any corrections or post-editing. Still, you can hover over the result in the TQI column to ensure there are no improvement suggestions. |
N/A | The LLM-translated target segment has been edited by a user, manually. In this case, neither the TQI result nor suggestions are displayed. |
No value | The AI Translate & TQI feature has been activated in a project and is set to work After analysis, but no segments have been LLM-translated during analysis (for example, because all segments had ICE matches/fuzzy matches and so were not sent for translation). In that case, when XTM Workbench is opened, the IS column is displayed. However, it does not contain any scores. |
These categories affect the final TQI result:
Translation consistency – assesses the translation’s quality, using the Multidimensional Quality Metrics (MQM) framework, evaluating how closely it resembles human-level accuracy.
Structural integrity – verifies adherence to formatting, glossary use, and structural elements such as inline tags, to ensure technical correctness.
Semantic quality – measures similarity between translations generated by multiple LLMs, assessing alignment, context clarity, and manageable complexity.
When AI Translate & TQI is activated in a project, LLM-generated translations are inserted in target-language segments, and the TQI results are displayed in the TQI column. If a Project Manager then deactivates AI Translate & TQI for the project in the Project Editor (for details, see Project Editor – General information: Machine translation), the TQI column and the TQI results in the Matches docked panel will no longer be displayed. However, the AI-generated translations will be retained in the target segments.