May 8, 2009 01:00
Contextual information in software localization
As you know I am supporting the Tanzanian Linux User Group (tzLUG) in their effort to localize Firefox 3.5. One of the challenges that the team faces is to have enough contextual information to properly translate the application. It is difficult to translate something that you have never used before, functionalities that the translation team is not familiar with or English words that can be understood in several ways... For example, the word "archive" could be a verb (to archive) or a noun (as in an -archive- of pictures). In Swahili the world will have to be translated differently if it is a verb (jalidi) or a noun (jalada). Without the contextual information, translations are normally of bad quality.
Having contextual information is needed to have good localizations, the challenge is not only to master the language but also to understand what exactly is being localized. A few years ago, while localizing OpenOffice.org to Swahili, we decided that it was good idea to add tags to the strings, so when the software was built we could backtrace bad translations. The concept is very simple, localize the software, build it and if a localized string does not make any sense when testing the software, use the tag to find the file where to change the translation.
If you like the idea, have a look to the The Translate Toolkit and podebug.