Localization of text to multiple target languages has always presented unique challenges. Although it is easy to translate a single version of a document into different languages, it is much more difficult to maintain translations over multiple versions of a document. First you need to find out exactly what has changed, then find the equivalent places in the translated document and only then begin the update.

With the move to representing documents as structured XML comes a new set of technical solutions based around intelligent three-way structured merging - synchronization. This paper presents a comprehensive and generic solution to the problems of managing changing "source language" documents and propagating those changes to translated versions in multiple languages.

By processing the original source-language document, the original translated version and the updated source-language text, we demonstrate how a "template" document can be generated which is very much easier to process for re-translation. This merged document has the original translation for any unchanged elements, and a template for any modified or new sections. By identifying precisely what needs attention and possible re-translation, and providing source-language and original-translation versions of the text, re-translation of complex documentation sets managed by large teams becomes practicable. End-user interactive editing tools can be built which simply process the resulting XML document.

We discuss the technical advantages of using Unified Deltas to archive files as well as the cost reductions and reduced time to market made possible by giving the translation team what they want - clear and simple indications of what has changed with "in situ" display of context. By identifying both what has changed and the exact location of the change, our technique allows translators to concentrate on what they do best - translating. Updating translations becomes less expensive, less prone to error, and a lot less tedious.

We present examples from real-world translations, and discuss extensions to the basic scheme outlined here. We show how this work builds upon existing intelligent XML change control software, allowing merging of multiple XML documents into a single document as well as tracking of changes in document structure and content. This technique offers a revolution in text translation which should be of interest to anyone involved in test translation - all made possible by the adoption of XML.