Shifting Translation Into HighBy John W. Verity | Posted 2005-08-04 Email Print
Who cares about owner's manuals? The German automaker does. It sees that book in the glove compartment as a means of improving its relationships with customers.Gear">
The biggest contribution of the new system is how it organizes the production process around a collection of reusable components. These include the graphical layout of each document, as well as the many pieces of text and graphics that flow into those layouts as manuals are prepared in different languages and for different cars.
In the past, Rüger explains, Volkswagen had used authoring software, supplied by Interleaf, that provided a WYSIWYG (what-you-see-is-what-you-get) view of a manual's pages. Authors could see on their screens exactly how a document would look, including type fonts, line spacing, word wraps and pagination. Each document was a hand-crafted product, created independently of others. But as the company sought to deliver manuals for more kinds of cars sold in additional geographic markets, it decided to adopt a more structured approach.
By breaking down manuals into hierarchies of precisely defined componentsbinders, booklets, sections, subsections and graphics, for instanceit becomes easier to share pieces of content across different car models and languages. Volkswagen opted to base its production process on standard generalized markup language (SGML), which can define the structure of a document down to any level of detail required. Once content is marked with the proper SGML tags, VW's editorial system can automatically search, retrieve, update and re-compose even multi-hundred-page documents full of interrelated drawings, photos and references embedded in text. "SGML has given us a lot of improvements in all of our publishing processes," Rüger says.
Translating the text in those components from German to another language remains a challenge, but component technology helps here, too. Volkswagen still depends on humans, not computers, to do its translations, as no computer has the smarts to autonomously turn lucid, stylish German into, say, its French equivalent. But computers can help translators do their job, especially with an owner's or service manual, which tends to contain straightforward wording and well-defined technical terms instead of clever metaphors and colorful figures of speech.
Trados' contribution to Volkswagen's new system helps translators reuse earlier work instead of having to translate each chunk of text from scratch. The software stores and manages Volkswagen's previous translations, from individual termssteering column, oil filter, etc.to complete sentences and even full paragraphs. Each of these translated snippets is linked to the corresponding original item, in German. In some cases, the previous translation may be usable as is.
If no perfectly matching sentence is available, a search will find text that includes the same keywords, structure and sense, thereby making the translator's job considerably easier.
While Volkswagen's earlier system included a database for storing text components, it did not provide the same search functions. With the Trados setup, Rüger says, "authoring becomes more like searching than writing." And through more-efficient reuse of earlier material, "we pay less." If a new car model uses the same engine as a previous model, the company may be able to reuse as much as 80% of the earlier car's owner's manual. If a new engine is employed, the reuse factor slips to about 50%, which still provides a significant savings over writing and translating the manual from scratch.
Alison Crawford, an IDC research analyst who has followed other deployments, says software can generally provide an exact match for 10% to 20% of all words, and a close, or fuzzy, match for 20% to 30%. The longer a glossary is in place, the better those percentages become.
And, instead of taking weeks to translate a manual, marketing materials or a Web site from one language into another, it can take only days, she says.