Nine cues to assess your language provider’s DITA-savviness

Nine cues to assess your language provider’s DITA-savviness

[DITA Loc Wire 21] As in mountain climbing, your language provider is a partner you trust, someone who takes your content further. When you adopt DITA for your content management, your language partner may not be fit for the technology leap. While your initial intent was to optimize your processes and lower your localization fees, you end up with longer lead times and unexpected costs. To ensure you are partnering with a DITA-savvy language partner who is fit for the climb, here are nine cues to look out for.

1. You pay for too many words

Your word count is inflated by 5 to 10% because content that should not be translated is not filtered. It could be code or address blocks for example. The excess word count can be difficult to spot when translation instructions are correctly enforced and the content is not corrupt

2. You spot duplicated index entries or indexes remain untranslated

Although indexes are used less and less since PDFs are being replaced by web portals for content publishing, they often indicate a translation process that has got out of hand. The culprit is the ill-managed Indexterm element, which is either closed to translation or not translated consistently.

Ideally, indexterms should be authored in a single reference file (conkeyref). Otherwise these indexterms should be in a term base (see the keyterm feature of WhP)

3. Your content is not properly indexed by the search engines

While you may have enriched your content with internal metadata such as Keyword to enhance searchability in English, your LSP may not have carried this over to the target languages.
The tagged elements should be in a termbase (see the keyterm feature of WhP)

4. Corrected errors still keep occuring

Your reviewers or your customer spot translation errors. You get them corrected. Nonetheless they reoccur in the next translation project.
This often occurs because the in-context review takes place when the content is published in the CMS. Errors were saved in the translation memories and are restored. There are many variants of this issue. This brought us to design the Augmented Review.

5. Linguistic review is fastidious

If you are working in a regulated industry, your certification body is likely to require a full in-context final review of a translated document, otherwise called Linguistic Review.
Bilingual segment reviews either in RTF format or online no longer work with the DITA standard. This will lead to a fastidious process with several steps:

  • The translated content is delivered and published.
  • Each release triggers a full review of the content, including parts previously reviewed.
  • Changes are entered manually in the CMS by the technical writer.
  • The same changes are entered manually by the translator in the Translation Memories.
  • If conditional content is implemented, this process might be applied to several publications to ensure all content is reviewed.

For example, one of our clients in life sciences reviews six documents for each translation.

WhP offers a solution that enables to review the content in context in the translation process, the Augmented Review.

6. Documents do not publish

Translated content delivered by the language vendor is either rejected by the CCMS or does not publish. Troubleshooting requires identifying the faulty file among the many files for a document and identifying the error. You then need to ask the language partner to correct the file, update the translation memory, and re-deliver the corrected files. Needless to say, this process is manual and time-consuming. This happens when translators are not fully familiar with DITA tags or when Machine Translation is applied without post-edit. We have seen document managers spend several nights troubleshooting translated content at every new publication.

With a non-DITA-savvy LSP, this might happen for one out of eight to ten projects. A DITA-savvy LSP will perform a full compliance check with the DTDs to detect these issues prior to delivery. They should also select and train DITA-savvy translators or have special expertise in applying Machine Translation to tagged content.

7. Your localization process breaks when a new technical writer is onboarded

You hire a new DITA-proficient technical writer and a few weeks later you obtain untranslated content, poorly translated content, or content that does not publish in the target language.

Actually, the new technical writer might have picked up a new element from the complex DITA grammar (610 elements + many attributes + elements from specialization). 

If the filter is not fully DITA-compliant, just like the so-called standard DITA filters from most CAT tools or TMS, it will break content, hide content from translation, or even suppress the elements.

8. Your translation vendor’s learning curve is too slow

You have migrated to DITA and are still in the transition phase, with limited availability and little DITA proficiency. You need to run many checks and iterations with your language vendor and teach him about DITA. We have seen a global organization take over six months to have their first DITA project translated correctly – and later opt for our services.
Choosing a DITA-savvy LSP enables you to benchmark and learn best practices for localization and source content creation.

9. You fail to reach your ROI target

The ROI of implementing or upgrading DITA often relies on reducing the translation costs, as you get rid of DTP and increase content reuse. If your language service provider fails to apply DITA-specific translation practices, you risk falling short of reaching your ROI.

If you have adopted DITA for your global content, partnering with an LSP that is not DITA-savvy can lead you to miss your ROI objectives and handle too many exceptions. These exceptions will in turn generate additional costs with your vendor. So a DITA-savvy LSP will deliver higher quality content at a lower price. We have heard our clients report increasing their productivity by at least 20%.

Recent posts