Systems would enable PC language translations
LONDON – Nicholas Negroponte, founder of the MIT Media Lab, recently warned that we’d wake up in five years and say: ”Gee, I didn’t realize that Chinese was the second biggest language on the Internet … after Spanish.”
He’s mocking the idea that most of the net will always be in English, and if it isn’t, how much will we understand? Aston Fallen, an American who lives in Paris, has the answer. It’s been in development for more than 30 years, and it’s called MT, for ”machine translation”.
Late next month, Fallen will travel to CeBIT, the giant German computer exhibition, for the European launch of Systran’s suite of packaged retail programs. These will enable users with PCs to translate between many popular language pairs. The programs run on Windows 95/98 and NT and Unix systems.
It’s not the first time Fallen has done something like this: he launched the Globalink Power Translation system, before the company was taken over by speech-recognition specialists Lernout & Hauspie. But he thinks it will be easier this time. ”The Internet has made all the difference,” he says. ”It’s no longer ‘MT? How do you spell that?’ but ‘When can I get it?’ ”
Five or 10 years ago few people were interested in software that could accept text in one language and translate it into another. Systran’s main customers were the European Commission and the American intelligence services, plus a sprinkling of large corporations. The EC also financed MT research, and sponsored development projects covering languages that had to be added to its roster when new countries joined.
The globalization of information via the Internet has changed all that. Most companies have e-mail addresses and Web sites, and they have to cope with e-mail from all over the world. As Systran points out, ”the ‘e’ in e-business doesn’t stand for English.”
Ordinary users are also surfing foreign Web sites, or finding articles in foreign newspapers and magazines, or chatting on bulletin boards. That and the rapid growth of a truly World Wide Web is stimulating the need for instant translations, and MT can provide them in a few seconds.
MT won’t, of course, provide the quality you’d expect from a professional. However, it usually provides enough of the meaning to make the minimal effort worth while.
Systran, a family-owned company based in Soisy-sous-Montmorency in France, is just one of a handful of major companies that provides a free translation service online. Its Babelfish system is heavily used because it’s available at the AltaVista and Go/Infoseek search engines. Users can also get instant translations at the Systransoft Web site (www.systransoft.com) by pasting in the text or a Web page address. The Web is providing Systran with a new market, and new marketing opportunities, even though people don’t pay for the facility.
But as Fallen says: ”Free is the first step towards making money on the net. You get the user hooked on the idea that he can do this, and then he thinks about tailor-made dictionaries and links to his site, and maybe offering translations on his own site. All of that leads to making money.”
Customized dictionaries are a key element, and Fallen says dictionaries are still the most important part of the MT equation for tackling idioms like ”it’s raining cats and dogs”. PC users have seen massive improvements in things like speech recognition and computer graphics over the past decade, thanks to massive increases in computer power, but MT hasn’t improved very much.
Fallen says: ”Chips are still not powerful enough to capture the nuances. I’m not a technology guy, but the power you’d need is above 2,000MHz PC to make a real difference.”
However, the MT industry has benefited from the arrival of powerful PCs. For example, Fallen says the European Commission is now moving to PCs instead of using mainframes, as PCs provide more flexibility. In the future, he also sees machine translation moving to handheld computers, and Systran is now developing a system to run on Microsoft Windows CE machines.
”A French company came to us – a very recognizable name, with a manufacturing facility in Hong Kong – and said ”Can you do it for CE?” The answer was ‘Yes, we can’.” Fallen says that all Systran’s products are now being developed on the GNU/Linux clone of the Unix operating system, and they’ll look at porting it to anything.
One day, mobile phones might also become powerful enough to handle machine translation, and WAP, the Wireless Application Protocol, might provide a motive for using it. But ultimately, what phone users would really like is not text-to-text translation but something that translates speech-to-speech in real time.
Fallen reckons we already have the technology to do that, at some level, but nobody could afford it.
But if ever we do arrive at the solution proposed in Douglas Adams’ novel, ”The Hitch-Hiker’s Guide to the Galaxy” – put a Babelfish in your ear – it will probably be a microchip-based fish running machine translation software.
(Distributed by Scripps Howard News Service. For more Guardian news go to http://www.guardian.co.uk/)