By Amy Sitze

Talk to anyone who has translated computer-based training (CBT) or Web-based training (WBT) from one language into another, and you’ll hear the same piece of advice being shouted from the rooftops: Plan ahead. Do your homework at the beginning of the project, say those who have been there—not as you go along. Here’s what else experience has taught them.

Research doesn’t have to be costly
Doing your homework doesn’t necessarily mean spending millions on research, said John Dees, technology advisor at Houston-based Kitba Consulting Services Inc. There’s a wealth of information about cultural and linguistic issues on the Internet, he said. While developing a CBT course for a Venezuelan oil company, for example, Dees found a university-sponsored Web site that provided helpful information on “Spanglish” computer terminology.
This is the second in a series of articles about translating training. Click here to read the first article, which gives an overview of the process. Both articles originally appeared in the February 2000 issue of Inside Technology Training and appear on TechRepublic under a special arrangement with the publisher.

Define your relationships up front
When you’re working with translators, make no assumptions about communications, said Kristina Nemeth, project manager at Wilson Learning Corp.’s e-Learning Division in Emeryville, CA. What language will you communicate in? How will you document your communication? Who on the client’s team is proficient in English? Who on your team is fluent in the client’s language? Who will be the liaison for phone calls and reviews?

Don’t assume text translation is easy
Most people don’t think about the amount of space different languages take up on the screen, said Wendy Collins, director of design and development at Raymond Karsan Associates in Princeton, NJ. If you’re translating English text into German, for example, there’s no guarantee that you’ll get a three-line to three-line translation. You may have to reorganize the layout and graphics to accommodate different word and sentence lengths.

Keep it simple
The more complex your initial script is, the harder it is to translate. Keep your instructions—both audio and text—as simple as possible, said Lisa Gatti-Dunn, lead instructional designer at Wilson e-Learning. Avoid colloquialisms that wouldn’t make sense in another language—for example, “It’s raining cats and dogs” or “Here comes the big cheese.”

Avoid “versionitis”
Don’t try to speed up the script-writing process by sending an early draft to the translators and assuming you can update it later, said Gatti-Dunn. Give translators the final draft so you don’t have three or four earlier versions floating around.
Have you worked on any translation projects? Is it easy to find translators? Have you tried to do any of it yourself? Send us a note explaining your experiences.
Now we’re talking
We’ve all heard stories about off-the-shelf translation software spewing out garbled misrepresentations of colloquial phrases. “The spirit is willing, but the flesh is weak,” for example, might come back in Russian as, “The vodka is good, but the meat is rotten.”

Translation software has its place, and it can be extremely useful for saving time and money. When the sentences you feed into the software are clear and correct, what you get back can be highly accurate. Many developers of WBT are finding that translation software is a good first step for translating simple text-based training into other languages.

When you move into spontaneous, conversational speech, however, it’s a different story.

“As human beings, we simply cannot get a simple sentence out without some corruption of the syntax,” said Alex Waibel, professor of computer science and director of the Interactive Systems Laboratories at Carnegie Mellon University in Pittsburgh. “We say, ‘uhhh’ and ‘umm,’ and we have false starts and stutters.”

Translators provide context
Human translators, who communicate the basic concept behind the sentence instead of analyzing each word separately, are one solution to the problem. Waibel calls this conceptual, as opposed to literal, translation. His research, both at Carnegie Mellon and at the Universität Karlruhe in Germany, focuses on bringing this type of conceptual translation into the digital world.

As director of an international group called Consortium for Speech Translation Advanced Research (C-STAR), he has worked with a team of researchers to develop something known as the interlingua, which is an “intermediate language” that uses computer code to represent the most basic elements of speech.

For example, let’s say you wanted to book a room in Heidelberg, Germany, using conversational English. You might say something like, “Well, you know, it’s Tuesday and I really need to crash somewhere, and, uh, it’s for a couple of days in Heidelberg.” Where a literal translation might have you wrecking your car instead of sleeping in a hotel room, the interlingua can sift out the colloquial expressions and extract the essential nuggets in the sentence: when, where, and how long you’re staying.

Although the interlingua hasn’t been applied to technology-based training yet, Waibel sees interesting possibilities. For instance, an instructor in New York might give instructions delivered by live streaming video to locations throughout Europe, with each student hearing those instructions in his or her own language. And this isn’t a futuristic dream, Waibel said. The technology is ready for the transition from academia to the real world. In fact, he has a spin-off company that’s already looking for partners.

To translate conversational speech, the interlingua system must be customized to a specific domain—travel, for example, or computer repair. Over the next five years, C-STAR researchers will focus on removing the domain limitations, Waibel said. The consortium is also exploring different delivery methods. What about special glasses that allow you to see subtitles under someone’s face as that person speaks to you in a foreign language? Or technology that uses video manipulation to move people’s lips so they appear to be speaking your language?

Waibel calls those “wild ideas, off the deep end.” And while they may seem wild today, it’s possible that we might find ourselves walking around Heidelberg 10 years from now, wearing our subtitle glasses and looking for a place to crash.