Word transform repair (disabled by default)
Overview
| Description | Attempts to repair “untranslated” content. |
| Default | Disabled |
| Supported Attributes | target_languages
|
Details
The word transform repair seeks to identify content that should be translated as-is into the target segment. The current implementation is limited to identifying differences between the lookup segment and the TM match, and to determining whether the difference has been translated verbatim in the target segment. If so, it performs a substitution.
In the following example, the untranslated block is the text between
In the following example, the transform is not possible since the differing text in the source, more and pies.
Lookup Segment: There are more crazy pies. Source Segment: There are more amazing pies. Target Segment: Amazing pies are still there. Repaired Source Segment: There are more crazy pies. Repaired Target Segment: Crazy pies are still there.The two words from the lookup segment were found in the source segment in the same order, but the text between the two words in the source was different. The repair process determines whether a repair is possible. The differing text is
amazing, and since it exists in the target, the transformation to crazy (from the lookup) is done. This repair assumes that since the original text was untranslated in the target segment, the new replacement text does not have to be translated either. This may or may not be correct and it needs to be validated by the translator.
of the amazing, is not found completely in the target. As a result, no transform is done.
Lookup Segment: There are more crazy pies. Source Segment: There are more of the amazing pies. Target Segment: Amazing pies are still there. Repaired Source Segment: There are more of the amazing pies. Repaired Target Segment: Amazing pies are still there.