LQA Assessments (TMS)

Content is machine translated from English by Phrase Language AI.

Available for

Team, Professional, Business, Enterprise and Software UI/UX plans

Get in touch with Sales for licensing questions.

Available for

Ultimate and Enterprise plans (Legacy)

Get in touch with Sales for licensing questions.

Language Quality Assessment (LQA) is performed by reviewers in the CAT editor to evaluate the quality of translated content based on pre-defined criteria. Issues are flagged segment by segment, categorized by type and severity, and stored in a scorecard that contributes to a final quality score.

While Quality Assurance (QA) checks are automated and look for technical issues like missing tags or typos, LQA assessments are performed by humans and focus on linguistic quality.

Creating and viewing LQA assessments are only available in workflow steps in which LQA is enabled. The LQA pane itself is present in the editor for all workflow steps.

LQA for an LQA enabled step can be enabled only during project creation or when adding the step to a project. If creating a project with an LQA enabled step, but turning it off during project creation, it is not possible to enable it afterwards.

When assessing content quality, different rules may be required for different content types or clients. Some content may need to have precise terminology and others may need to follow locale convention or be precise with target length and markup. Errors may also require different severities or assigned weights.

Overall feedback cannot be provided for joined jobs.

Add an LQA Issue

To add an LQA issue, follow these steps:

In the CAT Editor, select the LQA pane and click Start LQA.
Select a segment.
- To assess a particular piece of text, select the text inside the segment.
Click the flag icon next to the segment to open the Error category drop-down list.
Select an Error category, Severity, if the issue is a repeated one and an optional description.

Optionally, type or select @ to add a mention in the description and select the desired user or group of users from the list. Mentioned user(s) receive an in-app notification and email notification (if enabled in the project).
Click Flag issue.

Once an issue is flagged, it can be edited, replied to or deleted from the menu.
- Replying to an LQA issue allows adding a mention by typing or clicking @ and selecting the desired user or group of users from the list. Mentioned user(s) receive an in-app notification and email notification (if enabled in the project).
Repeat the last two steps to flag more issues as needed.
Click Finish LQA.

The LQA results and all flagged issues are listed in the LQA tab.

Segments with LQA issues are identified with a solid flag icon . Created LQA issues can be edited or deleted from the ellipses menu of a specific issue (not available for issues created by others).

Add Bulk Issues

Issues can be added to a selection of segments at the same time with a limit of 500 segments. The initial issue is considered the parent in a selection with the following issues being the children Child Comment . Editing or deleting these selections can then be performed in bulk operations instead of on individual issues.

Relations in bulked issues

Editing or replying to a child issue severs the relation to the parent issue (bulk operations can no longer be performed).
Replying to a parent issue does not break the relation to the children and replies are not propagated to the children.
Edits to a parent can be optionally applied to only the parent or to the children as well.
Deleting a parent severs the relation between the children (bulk operations can no longer be performed).

To add bulk issues, follow these steps:

With the LQA pane displayed, select a number of segments for flagging with an issue.

The selected segments' numbers will be indicated in the New issue window.
Select an Error category, Severity, if the issue is a repeated one and optionally provide a description.
Select how the issue should be flagged in the selected segments:
- Flag issue in subsequent segments as repeated issue.
- Flag issue in all segments as non-repeated issue.
  
  Repeated issues do not have this option.
Click Flag issue.

The issue is added to the specified segments with the parent issue indicated with and the children indicated with . Click the child icon to be taken to the parent issue.

Listed issues can be filtered Issue Filter . Options can be combined and include filtering by severity or Only repeated issues and Only parent issues.

Scoring of bulked assessments

	Repeated issue unchecked with Flag issue in subsequent segments as a repeated issue.	Repeated issue unchecked with Flag issue in all segments as non-repeated issue.	Repeated issue checked with Flag issue in subsequent segments as a repeated issue.
Segment 1	Unique error - penalty added (parent issue)	Unique error - penalty added	Repeated error - no penalty added
Segment 2	Repeated error - no penalty added	Unique error - penalty added	Repeated error - no penalty added
Segment 3	Repeated error - no penalty added	Unique error - penalty added	Repeated error - no penalty added

Generate LQA Assessments

To generate the LQA assessment, click Finish in the LQA tab. Overall feedback can be optionally applied.

A score will be calculated and a result will be displayed based on the pass/fail threshold set in the LQA profile. The score and the result are displayed in the LQA tab and on the jobs table in the LQA score column if LQA is enabled for that workflow step.

Tip

If the LQA score column is not visible, click the icon on the jobs table and select Customize columns to enable it.

The score is calculated from the job's total word count.

Scoring example:

A text of 500 words is set to fail LQA when it contains 1 critical Accuracy error and 1 minor Grammar error. The Accuracy error is given a weight of 2.0 and Grammar a weight of 1.0. Critical severity is penalized with 10 points and minor severity with 1 penalty point.

Count the number of issues for each error category and multiply by respective severity weight and error weight:
- Accuracy penalties = number of critical errors x critical error penalty x accuracy weight = 1 x 10 x 2.0 = 20 penalty points
- Grammar penalties = number of minor errors x minor error penalty x grammar weight = 1 x 1 x 1.0 = 1 penalty point
Add up the penalties:
- Penalty total = Accuracy penalties + Grammar penalties = 20 + 1 = 21
Calculate the pass/fail score, using the MQM scoring model [Score = 1 - Penalty total/Wordcount]:
- Score = 1 - Penalty total/Wordcount = 1 - 21/500 = 0.958
Present the score result as a percentage:
- Score % = 0.958 x 100 = 95.8
Enter Score % as your pass/fail threshold:
- Your pass/fail threshold is 95.8 %

Clicking on the results in the jobs table opens a scorecard window with options to send the scorecard by email or export as a spreadsheet (.XLXS) file. The link to download the LQA scorecard received via email expires after 5 days.

By default, the linguist assigned to the job can download their own LQA scorecard. To restrict scorecard downloads to project managers and administrators only, disable Allow linguists to download the LQA Scorecard in Access and security section of project settings. This can also be set as an organization-wide default in Access and Security section of Settings, which new projects will inherit.

The translation provider recorded on an LQA assessment (and shown in Analytics) is captured once, when the assessment starts, from whichever provider is assigned as translator on the previous workflow step at that time. If no provider is assigned yet, the reviewer's name is used instead. Reassigning the translator afterward does not update this value on assessments that already exist.

Note

Sharing LQA scorecard download URLs across different organizations (i.e. Buyer and Vendor organizations) is not possible.

Right-click on more than one selected job to have the option to download multiple scorecards.

LQA Score in Workflow Steps

When LQA is enabled in a workflow, all steps display the LQA score.

If there is only one LQA step, all workflow steps show the same score.
If there are multiple LQA steps, the score shown depends on the most recent LQA step:
- Steps before and including an LQA step display that step’s score.
- Steps after a later LQA step display its updated score.

Example

In a 5-step workflow where LQA is enabled in step 2 and step 4:

Steps 1 and 2 show the LQA score from step 2.
Steps 3, 4, and 5 show the LQA score from step 4.

As additional LQA steps are completed, the displayed score updates accordingly.

Assess a Repetition

Assessments can be added to repetitions with different flagging options with a limit of 500 segments.

To assess a repetition, follow these steps:

With the LQA pane displayed, select a repetition for flagging with an issue.

Repetitions following the selected one can be flagged together.
Select an Error category, Severity, if the issue is a repeated one and optionally provide a description.
Select how the repetition is to be flagged:
- Flag issue in subsequent repetitions as repeated issue
- Flag issue in all repetitions as non-repeated issue
  
  If the issue has been indicated as a repeated one, this option is not available.
- Flag issue in this segment only
  
  Subsequent or previous repetitions can be flagged independently.
Click Flag issue.

The issue is added to the specified segments with the parent issue indicated with and the children indicated with . Click the child icon to be taken to the parent issue.

Listed issues can be filtered Issue Filter . Options can be combined and include filtering by severity or Only repeated issues and Only parent issues.

Editing or deleting the parent issue gives options for whether the command should be applied to only the parent issue or to all issues created in bulk, including the children. Editing or deleting a child issue only applies the command to the selected issue.

Repeating LQA issues may require different treatment and may be:

Flagged as repeating and excepting the parent issue do not add any penalty points to the final LQA score.
Flagged as a unique issues adding penalty points to the final LQA score with subsequent issues being unique as well and reflected in the score.

Scoring of repeated segments

	Repeated issue unchecked with Flag issue in subsequent segments as a repeated issue.	Repeated issue unchecked with Flag issue in all segments as non-repeated issue.	Repeated issue unchecked with Flag issue in this segment only	Repeated issue checked with Flag issue in subsequent segments as a repeated issue.	Repeated issue checked with Flag issue in this segment only
Repetition 1	Unique error - penalty added (parent issue)	Unique error - penalty added	Unique error - penalty added	Repeated error - no penalty added	Repeated error - no penalty added
Repetition 2	Repeated error - no penalty added	Unique error - penalty added	No error logged	Repeated error - no penalty added	No error logged
Repetition 3	Repeated error - no penalty added	Unique error - penalty added	No error logged	Repeated error - no penalty added	No error logged