Replatforming with localization in mind

A wise woman recently said, “replatforming structured content is annoying and expensive.” This is doubly so when it comes to localization.

Replatforming nearly always involves content change—the new system may store content differently or require a different format or structure. Although the changes may affect your existing localization process, some of these changes may be for the better.

Content change and translation cost

Moving content from one platform to another usually requires reformatting or restructuring. Every system works with content a bit differently: linking, reuse, metadata, and storage formats may change even if the source content itself doesn’t.

Also, older systems tend to be littered with workarounds, proprietary tagging, and generations of iterative content adjustments. Moving content to a new system offers an opportunity to clean up and update existing content.

Your localization workflow must account for these changes. If the content itself is restructured, segmentation rules—how content is broken down into smaller strings of text for translation—will need to change. You may also need to reconfigure your translation tools and train your translators to work with new file formats. Even if the text itself does not change, you may see a drop in the percentage of 100% matches against your existing translation memory. ICE (in context, exact) matches may disappear completely, particularly if content files are broken up into multiple smaller files.

In any replatforming effort, factor in a cost increase when translating the migrated content for the first time. The cost impact could be sizeable depending on the amount of content affected and the extent of content change. Once the content is translated and stored in translation memory, the new formats and structures will be available for future translation work.

Leverage content intelligence

Replatforming offers a chance to bake more intelligence into how content is tagged, stored, and used. If reuse has historically been suboptimal (or non-existent), now is the time to leverage it. Work with your information architect to thoroughly plan out content reuse that is optimized for translation and authoring. Ideally, the reusable content components should provide enough context to be translated cleanly on their own.

Conditional text, variables, and other such features need the same consideration. You must ensure that any text that might be inserted programmatically or be included or excluded with conditionals is handled in a translation-friendly manner. Likewise, linguists must be made aware that these features exist in your content and how to handle them during the translation process.

Finally, consider how you might be able to leverage metadata to support your localization workflows. Evaluate existing metadata for effectiveness in translation, and consider adding metadata that may be useful going forward. For example, it’s useful to provide an abstract or description for each file that explains the intended audience and purpose.

Systems and workflows

Moving to a new content platform affects how systems and workflows interact with content. Any existing API connections are tuned for the legacy system, so they will need to be re-created, and there is a chance that some systems may not be able to connect optimally or at all. Before committing to any system change, make sure that the candidate systems provide the connections you need.

In particular, ensure that you can connect to your localization management system and that you automate much of the localization process using workflows:

• Approve content for translation
• Compare source content with the translation memory
• Collect and distribute content for translation
• Report on completeness of the translation
• Designate reviewers and track review completeness and translation quality
• Stage translated content for publishing

Automated workflows let you speed up the entire translation cycle and free up bandwidth for quality control.

If you are replatforming your content and have concerns about localization, we can help.