Translation tech helps firms talk business round the world

A man uses Google's voice translate app on his smartphone during an event at the City of Fashion and Design, Paris

English is the language of business, right? Er... not really.

If the oft-heard maxim were universally true, the outsourced translation market wouldn't be worth a chunky $37bn (£23bn) or be growing at about 6% a year.

While content in English still dominates the web, "billions of people don't read English at all or well enough to make buying decisions," concludes a survey by Common Sense Advisory, a business research consultancy.

In reality, businesses must translate and localise products into a host of different languages, and that requires linguists.

And with competition among translation firms fierce, many are turning to technology to steal a march on their rivals.

Found in translation

Take Thebigword, a Leeds-based firm employing about 12,000 linguists in 73 countries covering 500 languages. Its previous clients include Ricoh, Electrolux and DHL.

The firm says it can connect you with an interpreter in just 30 seconds by phone, localise your market messages or translate your documents.

However, it believes the real weapon in its technological arsenal is its new "translation management system" (TMS), which claims to deliver projects four times faster than the industry standard by automating project workflow and using computer-assisted translation tools.

Leeds-based translation firm Thebigword employs 12,000 linguists worldwide

Imagine you are a technology business launching a new phone in multiple countries on the same day.

"Translation for a job like that goes through a huge amount of different processes - it's not just one page sent, translate it and send it back," says chief executive Larry Gould.

"It may have diagrams on it or illustrations, or need to be presented a certain way, localised, edited and double-proofed. And of course you've got to do it all to tight deadlines."

Mr Gould can use up to 200 linguists on a single project like this, spread across 33 countries. But the TMS helps co-ordinate the process, allocating the workload across time zones to speed things up and cut costs.

Language barrier?

Such firms still rely heavily on human linguists but are increasingly complementing them with lower-cost automated "machine translation" tools, known as MT in the business.

Ben Sargent, a senior analyst at Common Sense, says such technology has its drawbacks, but can work well for low-stakes, high-volume work, particularly online.

In the book of Genesis God introduced many languages to punish mankind for building the Tower of Babel

"We estimate that less than half of 1% of all the digital content that could and should be translated, actually is.

"No-one has the budget to do all that. So in low-value content applications, like user-generated content and consumer-to-consumer interactions, automated translation gets a lot more traction."

For example, when eBay realised that more than 20% of its sales involved cross-border trade - and that its international business was growing faster than its domestic business - it acquired AppTek's machine translation technology to help meet the demand for local language listings.

Currently eBay only translates listings in countries such as Brazil and Russia, but eventually wants to help sellers list their goods in multiple languages, and chat to buyers via instant messaging that translates in real time.

There are about 6,500 spoken languages in the world today

But Ryan Frankel, chief executive of VerbalizeIt, a translation firm that uses only human translators, is not convinced by MT.

He believes that it "is light years away" from delivering anything beyond a "get the gist" solution.

"Businesses rightly value accuracy but also brand, industry and cultural-specific terminology and nuances that require an experienced and trained community of translation professionals."

Even advocates of the technology admit that an accuracy rate of about 70% is considered excellent, but that this can only usually be achieved for technical documents using highly consistent terminology.

Once the slang phrases, idioms and metaphors of normal human conversation are thrown into the mix, accuracy can plummet to 30%.

Beam me up Scotty

"Speech-to-speech" technology, which translates the human voice in near real time into text or words spoken by your computer, offers some exciting possibilities.

Although it is certainly in its early stages - Microsoft's pre-beta Skype Translator tool, unveiled in May, was clunky and slow in demonstration to say the least - speech-to-speech is evolving fast.

According to analyst Gartner the market is likely to mature in the next five to 10 years.

SpeechTrans, a frontrunner in the space, claims its users can have a conversation in more than 40 languages over fixed-line or internet phone, and that the technology can even recognise different accents and dialects.

"In four years we have been able to accomplish more than was ever thought possible in this area of technology," says Mark Coviello, director of sales.

Misunderstandings can sometimes have serious consequences...

"At this rate we foresee the ability for any human being to communicate with any other human being without error [using speech-to-speech]."

Hewlett-Packard has already integrated SpeechTrans into MyRoom, its web conferencing platform, enabling business professionals "anywhere in the world to collaborate in 30 different languages, in the same conversation, at the same time".

The technology giant won't tell the BBC how many users it has, but says "adoption is increasing", along with the application's accuracy and speed.

'Dubious results'

Still, Mr Sargent says such platforms have a long way to go before they really take off.

"Progress can be slow, and the utility of these systems today is limited.

"Don't expect to see courtroom or hospital interpreters being replaced anytime soon, except for emergency situations, where dubious results may still be better than nothing."

Automatic machine translation services are not always that accurate, but they should improve as they learn

As with MT, some also doubt whether speech-to-speech will ever really grasp the nuances of language the way a human can. Still, Mr Coviello claims taking humans out of the equation has its benefits, too.

"We find that there may not always be a person available for translation when needed the most, or that due to certain ideologies or conditions the person translating may mistranslate what is being said."

While businesses will surely explore automated translation technology for its cost-saving potential, it seems that demand for high-quality human translation will continue to grow as multilingual content proliferates.

"We believe new technology has to be embraced because there is just so much communication required out there," says Thebigword's Mr Gould.

"It's brilliant for our industry - [technology] will enhance our business, not take away from it."