Translation and Treatment
by Sandra Guy, 1st June, 2005
Online services and computerised language translation systems are overcoming the Tower of Babel and helping doctors to overcome language barriers
Translation and treatment Amednews - American Medical Association By Sandra Guy 3 May 1999 Online services and computerized language translation systems are overcoming the Tower of Babel When Daniel Carlin, MD, e-mailed instructions to a Russian sailor that allowed the sailor to do emergency surgery at sea on his own infected elbow, it captured national news. One of the key elements that made this story possible was the online service the Boston emergency physician used to translate his words before he transmitted them. The sailor, Victor Yazikov, was on the first leg of the "Around Alone" race, about 400 miles off the coast of South Africa. Each sailor in the race has to have a laptop, linked via satellite, with an Internet connection. Another racer, Isabelle Autissier, asked that Dr. Carlin occasionally consult with her physician, Dr. Jean-Yves Chouve, a French native and a skilled sailor. Dr. Carlin, who practices telemedicine fulltime, now consults frequently with Dr. Chouve via e-mail by typing his message in English, cutting and pasting it into a clipboard, moving to AltaVista's translation site, pasting the message there and pushing the button that translates it into French. Years of experience with online translation services prove that the simpler the sentences, the better. Idioms pose the biggest hurdle because the translations are so literal, Dr. Carlin notes. For example, if he e-mails that one of the racers is "really up the creek," the translation will say the racer is in trouble because he is stuck on a river. Another problem involves words with multiple meanings. A phrase such as "got to go" becomes "get to produce." And the French word encore can mean again, not again, yet again or sometime real soon. "You have to set the stage for the use of the word," Dr. Carlin says. Putting a face with a name A language translation system being developed at Carnegie Mellon University, Pittsburgh, Pa., hopes to take online translation beyond written messages by enabling people who speak different languages to hear one another. Its most dramatic feature, however, is an avatar, or a visual representation of the user. Though avatars can take the form of an idealized person, an animal or an inanimate object, in this case they synthesize the faces of the speakers as they converse in a shared virtual reality. The avatar software is designed to mimic a speaker's facial expressions, says Monika Woszczyna, PhD, a visiting researcher from Germany active in the project. The system takes a still photo of the person speaking. Then it seeks out key features that show how a person looks when he or she talks, particularly around the corners of the eyes and the mouth. It also pinpoints eye width, as well as the location of the eyes and nose. The speaker then reads a sentence, and the system "learns" his or her speaking characteristics, such as whether he blinks before he speaks. A separate element will smooth out the idiosyncrasies so the avatar seems to be speaking naturally. The avatar's lip motion is synchronized with the synthesized speech. The limitation of this type of avatar is that it can do only what the speaker does. Other avatars start with a generic "wire-head" model, Dr. Woszczyna says. These systems take a picture of the speaker and pull it over the model like a mask. The wire-head model allows a wider variety of actions, but it cannot reproduce the speaker's unique characteristics. Studies show that a viewer who sees someone speaking understands the context of the remarks better than a listener who hears only the audio, says Jackie Fenn, vice president and research director of advanced technologies for GartnerGroup, Stamford, Conn. The challenge for avatar-based language-translation systems will be to overlay expressiveness onto an avatar so that it enhances the listener's understanding, she says. Lernout & Hauspie of Burlington, Mass., which makes dictation software for physicians, has its own prototype of the Holy Grail of language translation. Its demo features a businessman speaking English on the telephone, while a listener in China hears the man's voice speaking Chinese. The business model emerges The language translation tools in development are rudimentary compared with the avatar and real-time models. Machine translation has been available for decades, primarily for crude text renderings. Typically, the software assigns to each word a part of speech and parses the sentence. The program then can synthesize the foreign language version using the target language's basic grammar rules. The increasingly costly alternative to machine translation is to use a human translator from start to finish, which usually occurs with documents that contain idioms, sensitive information or specialized terminology. Yet futuristic solutions aren't pie in the sky. That's because language translation is fast becoming a hot Internet commodity. Expect medical applications to grow, too. Dr. Carlin, whose ocean practice serves business travellers, super yachts and professional boaters, admittedly is one of a few clinical distance medicine physicians. He runs a 24-hour virtual emergency department and, with a colleague, has started to provide virtual radiology consults to a nurse who transmits digital images from an Internet cafe in Africa. Yet Dr. Carlin foresees global connectivity among patients and physicians, as well as Web sites that will let clinicians work more closely with researchers. Such collaboration, he is quick to point out, has sparked medical breakthroughs throughout history. "The medium is now the Internet," he says.
©
Back to the Translation and the Computer
Back to the main articles index
