Translation Memory Repair Component
Repair components allow you to define how to correct differences between lookup text and source text, and how to correct the target.
- Identify the difference between the lookup text and the match source text to be repaired. For instance, perhaps they differ only by a number.
- Identify the match target text that corresponds to the match source text to be repaired. For instance, this could be the same number, or the number in a different form, such as a localized form.
- Repair the match source text. This is relatively straightforward. However, this is key. The scoring process is based on the difference between the lookup text and the repaired match source text.
- Repair the match target text. This is generally the most difficult part. Correcting numbers that are the same between the source and target text is straightforward, but few repairs will be so simple.
- Set the repair penalty. Depending on how effectively or reliably you are able to repair the target text, you may want to assign a repair penalty to signify the confidence level of the repair.
- Repair differences in words, numbers, dates, etc.
- Automate data conversions based on locales, such as converting from English measurement to the metric system.
- Localize numbers, dates, currencies, etc.
WorldServer supports a repair framework that allows individual repairs to be enabled or disabled for specific locale combinations. What this means is that repairs can be safely created to address issues in specific language combinations. The WorldServer system administrator can always disable specific repairs for language combinations as necessary or desired. However, the repair implementation can also monitor for itself the language combinations for which it should apply the repair strategy. As the implementor, you can choose to disallow the repair for various locale combinations, or you may even choose to penalize other combinations where the results may not be as reliable.
Java documentation
com.idiominc.wssdk.component.tm