WorldServer - How to set up BG4/MT Cloud Adapter in SDL WorldServer with the "Apply Machine Translation" step in a workflow so that MT populates all segments that are not 100% or ICE match?

Article Number:000015305 | Last Updated:6/11/2024 10:36 AM

Scope/Environment

WorldServer

Question

How to set up a Language Weaver Cloud Adapter in WorldServer with the Apply Machine Translation step in a workflow so that MT populates all segments that are not 100% or ICE match?

Answer

To get MT to populate all segments that are not 100% or ICE match, the following configuration is needed:

1) In the Management > Linguistic Tool Setup > Machine Translation Configurations, select the configuration
   a) Under Translation Configuration, do the following:
        Set Fuzzy Score of MT Segments to: 100
Do not translate matches above Score: 100

          WS MT configuration

To apply machine translation, you have 2 options:

2) Enable the option Leverage Assigned MT in the Segment Asset step by selecting Yes:

User-added image

2) In the workflow, in the Apply Machine Translation step
a) Set the Minimum pre-translation score to 100
WS Workflow

In both scenarios (Segment Asset or Apply Machine Translation) it is important to note that MT will not overwrite any 100% or ICE translations. MT will only get applied to blank target translations at the Apply Machine Translation step.

If you have a Pretranslate Segments step after Segment Assets (where MT was applied), this Pretranslate Segments step will remove all MT translations inserted during the Segment Asset step and pre-translate from the TM only.

If you have an Apply Machine Translation step after a Pretranslate Segments step, only segments that were not pre-translated from the TM will be machine-translated

If you do not have a Pretranslate Segments step (and Segment Assets does not apply Machine Translation), the Apply Machine Translation step will pre-translate all segments that are not 100% or ICE Match pre-translated during segmentation, including those where a fuzzy match from the TM is available. So if you want to make sure that fuzzy matches from the TM should not be overwritten by the MT, add a Pretranslate Segments step to your workflow before Apply Machine Translation

For live TM mode (in tm.properties, enable_live_translation_memory=true):
MT translations in live TM mode will have the status="None", for the translations to be updated in the TM, the status needs to be set as Reviewed or Pending Review

For non-live TM mode (in tm.properties, enable_live_translation_memory=false):
If the option is to not update the TM for the MT translations:

Do not select Save and Update TM Browser Workbench
Any Save step or other steps in the workflow after Apply Machine Translation should be configured to NOT update the TM. The Save target step never updates the TM so use this step if you want to make sure that when creating a target file, the TM is not updated.

Reference

Send Article Feedback