With the fast pace of technology, product updates are a frequent necessity that not many can escape. Even if you are planning to start a completely new localization project, it is important to understand the different process nuances because sooner or later, your product will require updates. Implementing a process up front that can effectively handle updates will save you time and money in the long term.
Bottom-Up Localization Process
The bottom-up process dates back to the early days of localization. It relies on simple diff software to compare old and new source files. Once the changes in the source files are identified, translators can apply them to the previously translated (target) files and perform all additional necessary translations or updates.
While this process works well when changes are minor, it becomes highly complicated and error-prone when the percent of changes starts exceeding single digits.
Also, translators’ time is used inefficiently because each translator will be making the same code, tag, and layout updates to the old translated files, for each and every required language, while they are translating the updates.
Top-Down Localization Process
With the advent of search engine and database technology, translation memory (TM) tools started emerging from companies such Trados, SDL, Atril, Star, and others.
TM products store source and target translation segments in a database. A search engine running in the background can detect the presence of an already translated segment in the database and offer it to the translator at any time in the translation process. This eliminates the need to re-translate any already translated segment.
While performing updates, the translator relies only on the new source files and the translation memory (TM) database. There is no longer a need for the old translated files. The search engine finds all previously translated strings stored in the TM database and prompts the translator to accept them into the new source files. At the same time, translators no longer need worry about updating, code, tags, or layout, outside the segments to translate (as they must in the bottom-up approach). This methodical work through the new files makes the process much more predictable and manageable, as well as much less error-prone. It also enables translators to focus on what they do best— translate!
Reusing translations from the TM for the entire file will require linguistic review, layout changes in a desktop publishing (DTP) program, and quality assurance (QA) work. These tasks can be delegated to the appropriate review, layout, and QA staffs.
Another side benefit of the top-down localization process is that it allows many translators working on the same project to share translations simultaneously via the TM database. This ability promotes both efficiency and consistency.
Better quality, more efficiency, and time savings – what more could you ask for?
Enter the hybrid top-down/bottom-up localization process.
GlobalVision International’s hybrid top-down/bottom-up process harnesses the best of both approaches. It is particularly useful when working with projects involving hundreds or thousands of files, such as online help or single-sourced manuals built using HTML or XML.
To use the hybrid localization process, you will need access to the old source files and the corresponding target (translated) files. The process uses an automated diff approach to separate the new source files into three categories: no changes; minor changes; and major changes/new.
No Changes: All unchanged files can be reused as-is from previous translations that were done and released, eliminating the need for translation leveraging, updating, layout changes in a DTP program, and QA work.
Minor Changes: Files that have changed slightly can be put through the bottom-up localization process effectively. This will again eliminate the need to leverage and “re-DTP” them. Language specialists will still need to make minor translation changes, and final QA is still advised.
Major Changes/New: New or heavily changed files can have the top-down process applied to them to avoid human errors and make use of translations stored in the TM. Two processes can be followed here:
- Standard: Segment-by-segment leveraging from the TM on all files, requiring linguistic review or check of each segment to ensure accuracy of the leveraging. Segment automatic leveraging is not recommended.
- XTranslate: A technology from Trados enabling paragraph leveraging before the segment leveraging process is applied. Paragraph automatic leveraging can be trusted to do away with the need to linguistically review or check leveraged paragraphs; leveraged segments, however, will still need to be linguistically checked. This technology is limited in the file formats it supports, and it requires the availability of previously translated “uncleaned” files. (Refer to Trados for more information.)
Once files are translated, layout changes and QA will be required on new and modified files.
The hybrid localization process is the best approach to follow. Its impact varies according to the volume and format of files, the frequency of required updates, and the sophistication of the tools and people applying it. It does take specialized tools and experienced localization professionals to understand its abilities and limitations, and to apply it where it can be most effective. Your savings in time and costs, however, can be truly significant!