You have to market to them in their own language
1 March 2001

Computerised systems that promised to deliver fast, cheap translations have never matched humans for accuracy. Machines that can't recognise metaphor and innuendo still need the personal touch, says Chloe Veltman

IN a brightly-lit seminar room at San Jose's pristine Entrepreneur Centre, 10 Californian artists are learning how to sell their work on the worldwide web.

The seminar leader, a local artist called Marques Vickers, shows the group his website, a tapestry of colourful paintings, sculptures and text.

Everyone is busy admiring Vickers' biography page when he clicks the mouse and the whole text suddenly appears in German.

"If you want to reach a worldwide audience, it's good to present yourself in a range of languages," he tells the group as they watch his text move from German to French.

Although English speakers were the creators and first users of the internet, Americanese, the erstwhile lingua franca of all things digital, has just been usurped by the growing non- English speaking web audience.

According to IDC, a technology research company, the English-speaking online population stands at 192.1m, compared with a non-English population of 211.3m.

By 2003, IDC estimates that the number of non-English users will have grown to 560m, dwarfing a English-speaking population of 230m.

Advertisements for Accenture (formerly Andersen Consulting) depict peaceful oriental scenes, predicting "Chinese to become the No 1 web language by 2007." The entire network economy, from the largest multi-national to the artist in his studio, is turning polyglot.

"The language barrier is the main impediment to globalisation right now," says Mike Dillinger, director of linguistic development at Logos Corporation, a trans-lation software company. Yet some companies, particularly in America, have been reluctant to accept the need for a multilingual business model.

"In the US, many businesses think English is the only language," says Bill Dunlap, chief executive of Global Reach, a Silicon Valley-based online marketing company.

Dunlap is quick to point out the mistake. He says: "Marketing always takes place in the language of the target market. It doesn't matter how conversant people are with English - if your target market lives in a country where English is not the native tongue, you have to market to them in their own language."

The idea is catching on, at least in the blue chip sector. Virtually all Fortune 500 companies from Sony to General Motors now build multilingual websites, software applications and documentation systems, employing specialist translators and a variety of translation technologies.

But the task of sustaining perfect, up-to-date parallel websites and multilingual business tools across a range of languages is becoming increasingly implausible. Human translators alone cannot meet the sheer volume and speed of the global economy's linguistic requirements.

"Can you imagine The Daily Telegraph translating five or 10 years of e-content into 30 languages?" asks Dimitrios Sabatakakis, the French chief executive of Systran Group, a well-known translation technology company. "The return on investment is impossible and the job would take several years."

Computers have been aiding translation for a long time. In fact, the concept behind machine translation (MT) predates computers by several hundred years - Rene Descartes suggested inventing a symbol that would stand for equivalent ideas in different tongues in 1629.

When computers came along in the middle of the last century, MT was one of the first computer applications designed to act on words rather than numbers, never mind e-mail and word-processing.

By the late 1950s, MT had become a thriving industry and computational linguistics was suddenly the sexiest field in academic research. Institutions such as the Massachusetts Institute of Technology and Carnegie Mellon University attracted vast sums in governmental and private sector support.

Since then the drive towards MT has garnered as much criticism as praise. Computerised systems that promised to deliver accurate translations at high speed and low cost never matched humans for accuracy.

Gabe Bokor, editor of Translation Journal, says: "Despite huge efforts by governments and corporations to develop usable computerised language translation tools, current offerings are still ridiculously primitive compared with the capabilities of even moderately-talented human translators."

Therefore most professional translation has been divided between humans and machines. Technologies becoming standard include Translation Memory (TM), which digests and automatically recalls standard phrases initially translated by a human.

Web-based models are operated by such companies as Logos Corporation, where customers submit texts over the web which are subject to machine translation, human revision and billing.

Dr Celia Rico, head of the translation and interpreting department at the Universidad Europa in Madrid, believes that human translators need to be technologically savvy to work in today's competitive market.

She says: "As the translation process changes to meet new market requirements, human translators need to evolve as well."

Humans play an important proof-reading, nuance-seeking and editing role in the translation process, with an emphasis on optimising the human translator's costly time through technology.

As Dillinger puts it: "We provide a system for relieving the translator of the drudgery of doing routine translation. This frees up his or her time for the more challenging parts of the task, such as metaphor and innuendo, that require human artistry."

Instead of reaching for the improbable goal of flawless human quality translation, technological advances are now focused on specific tasks.

An example is Silicon Valley-based research organisation SRI's partnership with Telia, a Swedish telecoms company, to create speech-to-speech English-Swedish translation for booking airline flights.

Steve Appleby, translation project manager for BT's Language Technology Group, says: "Since translation is such a complex area, different compromises are possible for different situations."

These compromises are making globalisation possible for the small business sector and the casual web user. The web offers some moderately useful free translation options for those of us who don't need to spend more than $2m creating what Global Reach's chief executive perceives as a "decent multilingual website".

Examples include AltaVista's Babel Fish translator, powered by software from Systran, which gained permission to use the name of the translating fish created by Douglas Adams in The Hitch Hiker's Guide To The Galaxy, and FreeTranslation.com, which uses Transparent Language software.

AltaVista says it receives 30m translation requests every month. It offers 19 language pairings, including Korean to English and English to Portuguese, as well as a Virtual Keyboard powered by SlangSoft, which allows users to type text in seven different languages.

Translating "Shall I compare thee to a summer's day?" into French produces quirky results: "Est-ce que je dois comparer le thee au jour d'un ete?" But, as Matt Costamagna of AltaVista says: "Babel Fish translates best the concise grammatical style found in newspapers."

In his best-selling book, The Age of Spiritual Machines, Ray Kurzweil is optimistic about the power of computers to equal and even surpass human linguistic capabilities.

He wrote: "In the next decade, we will see translating telephones that will provide real-time speech translation from one human language to another."

But the father of the Babel Fish, Douglas Adams, is less certain about the speed of evolution.

He says: "Until computers came along, the only entities we'd ever had to teach language to were ourselves, and we already had an innate ability.

"It was only once we tried to teach a computer how to do language that we began to glimpse the dizzying scale of it."

He asks: "Will we ever get there? We might eventually, but I think we are going to spend an awfully long time in an apparent Xeno's Paradox."

This paradox, for anyone not familiar with it, asks how you ever reach somewhere if you keep moving half the distance towards your goal. It's a question many of us have asked when starting to learn a new language.

Let's twist again

By day, Michael Reck is a manager at IBM Global Services in Stuttgart, Germany, but by night, he is the man behind the First International Collection of Tongue Twisters, a website crammed with 1846 sound bites in 77 languages.

Reck's obsession with tongue twisters began in 1984 during a party with people from more than 12 countries who spoke more than 15 languages between them. The group began to exchange tongue twisters from their own countries, and Reck has been collecting the rhymes ever since.

The collection was first launched on the internet in 1995 while Reck was a student at Innsbruck University. In January his website averaged 1,354 hits per day with visitors spluttering over such gems as "a aha ya ha a ha ya a" (meaning "I went, I played, I fell" in Guarani) and the testy-sounding Wolof rhyme, "tuki fuki buki gudi tuki fuki buki bechek".

One of Reck's favourite specimens of English tongue-twisterism is: "If two witches would watch two watches, which witch would watch which watch?"

If that was a doddle, try saying: "the sixth sick sheik's sixth sheep's sick," as you're running for the bus.

Copright The Telegraph Group Ltd