The translation repair SDK component allows you to implement custom methods for resolving differences between a source lookup segment and a source segment stored in the translation memory.
During the lookup process, if attempts to find ICE matches and exact matches fail, repairs are attempted to improve match results. Often, differences between a source lookup segment and a source segment stored in the translation memory can easily be repaired. In some cases, these differences can be automatically repaired, minimizing the work required of the translators.
Repairs are in-memory changes; they are never automatically stored in translation memory translation records. If the translator agrees to the repairs, the translator can choose to add new translation records to the translation memory when the asset being translated is saved.
The repair process may or may not be able to repair segments to 100% levels. If the initial unrepaired segment score was below the configured threshold, then WorldServer translation memory will never allow it to be scored up to 100%. Nonetheless, for segments where the initial differences are minor, the repair process may be able to repair many of the segments, increasing the ability to leverage the translation memory, and decreasing translation costs.