What is LQA?
The LQA (Language Quality Assurance) feature is a translation quality scoring system that is based on the Multidimensional Quality Metrics (MQM) model which has been designed as part of the QT Launchpad project. For more information about MQM, see: MQM.
We recommend that you read this article, which contains a step-by-step guideline to activating and performing the LQA step: How to perform LQA.
As a user with the Administrator role, go to Configuration → Settings → LQA → LQA, where you can select the full list of issues or any subset of them. You can also set the weight for each item and severity multiplier for minor, major and critical items:
Weight: The importance that the administrator would like to assign to a specific error.
IMPORTANT!
It is advisable not to apply any weight that is higher than 1.
Severity: Another way to change the importance of a specific error.
You can also create and configure your own LQA errors to be applied in XTM Cloud projects.
Apart from that, you can always specify if you would like to base your LQA score on the source or target word count in Configuration → Settings → LQA → LQA Options:
LQA calculation method
The LQA score is calculated as follows:
Penalty (P) score is calculated according to this formula:
P = ( Issuesminor × Minor multiplier + Issuesmajor × Major multiplier + Issuescritical × Critical multiplier )
Adjusted Penalty (AP) score is calculated according to this formula:
AP = P × Issue Type Weight
Total (TS) score is calculated according to this formula:
TS = 1 - ( AP / source or target Word Count )
Example:
a) weight of all LQA errors = 1; b) severity for Minor = 1, and for Major = 5. Penalty (P) score for this project is:
Adjusted Penalty (AP) score for this project is:
Total (TS) score for this project is:
|
---|
How does LQA differ from QA?
XTM Cloud can perform different types of error checking. QA checks are performed automatically in XTM Workbench, on the basis of a QA profile that is applied to the project. LQA, however, is always performed manually, by a human linguist.