A bid to spur research into new translation technologies could revitalize an industry that has been laboring under a human-vs.-machine split for decades
Editor’s Note: An earlier version of this story incorrectly stated the number of words that translation services can convert per day.
For decades, machine translation has been the next big thing. With every tiny advance, companies and researchers predicted that speedy, accurate language translation, completed wholly by computers, was just around the corner. But the technology has never quite caught up, and the promise of a global market free of language barriers has yet to materialize.
But there is progress. Companies have combined the power of humans and computers to simultaneously double the speed of translation and nearly halve its cost. Where each translator once converted 2,500 words a day at a cost of some 25¢ per word, they can now offer 5,000 words a day at around 12¢-15¢ a word. The savings add up mightily when a project can, for example, involve several million words. Still, the amount of information generated in the Internet Age represents a deluge. Software has progressed, but language changes frequently and begs multiple interpretations. Even with today's most cutting-edge technology, there are more words to be translated than most companies or governments could ever afford to handle. This shortfall limits opportunities for companies to market and support their products across languages, and to conduct business on a global scale.
Now, however, a direct challenge from the Obama Administration to achieve accurate, real-time translation of major languages—a challenge that comes with cash-for-research as an incentive—could spark new technologies and erode the language barriers that still hamper international business. As detailed in a Sept. 20 white paper from the White House, some $1 billion of the $787 billion stimulus package will go to such innovation projects. The effort is being touted as part of the Administration's push to reinvigorate science and technology innovation in the classroom and workplace.
White House's Translation Goal
In the White House paper, A Strategy for American Innovation: Driving Towards Sustainable Growth and Quality Jobs, the Administration lays out what it expects from the $1 billion federal investment in innovation. Amid new energy, health-care, and education goals, the report offers several specific "grand challenges" for the 21st century. Among them: "Automatic, highly accurate and real-time translation between the major languages of the world—greatly lowering the barriers to international commerce and collaboration."
While it is not clear how much of the $1 billion in stimulus money will go directly toward translation efforts, any federal investment could go far in the industry. The machine translation industry hasn't grown much, hovering around $100 million for years, and the recession has only exacerbated the situation, leaving plenty of linguists looking for work, says Don DePalma, chief research officer at Common Sense Advisory, a Lowell (Mass.)-based translation consultancy. "There's a lot of pent-up intellectual capacity that could really improve natural language processing," he says.
For global companies, the benefits of better, cheaper translation services are obvious. In a globalized marketplace, companies need to advertise their products in multiple languages. They need to drive traffic to their Web sites, especially where commerce requires it. And if companies want to have a respected, trusted global brand, they must provide support for their non-English-speaking customers. What's more, customers perceive companies that speak to them in their native tongues as more credible, DePalma says.
Helping Companies Reach Customers
In fact, 52% of consumers will only buy something from a Web site in their own language, according to a 2006 Common Sense report that surveyed more than 2,400 consumers in eight countries. In France and Japan, that figure increased to more than 60%. Consumers who did not speak any English were six times more likely to avoid English Web sites altogether. What's more, 64% of those surveyed said they would pay more for a product if they could get information about it they could read. "When you're dealing with anything really expensive or that potentially involves a long-term financial decision—like life insurance or stocks—customers prefer to have information in their own language," DePalma says.
The role of translation is not restricted to reaching customers—businesses typically find savings by using the language of countries where they operate. And in a recession, trimming costs is paramount. For example, if a consumer can read a company's manuals on its products or services on the company Web site, that person often will not need to call a customer service center—a major expense for most companies.
Microsoft (MSFT) recently used machine translation for the release of its Microsoft SQL Server 2008, a database management system, which increased the time to market but reduced the project cost. The software was released simultaneously in 11 languages and costs decreased by up to 6% per language with 7 million to 9 million words translated in each language, the company says. Microsoft also plans to release its SQL Server documentation in Turkish for the first time. Dell (DELL), Intel (INTC), and General Motors all say they're also using machine translation.
Human-Assisted Machine Translation
The most advanced translation providers—firms such as Madrid-based Linguaserve, U.K.-based SDL (SDL.L), and Lionbridge Technologies (LIOX), based in Waltham, Mass.—use human-assisted machine translation (HAMT). With HAMT systems, text is fed into a computer program that tackles the first round of word and sentence conversion using statistics, language rules, or matching with past translations. That covers about 90% of the work. A human then steps in to correct mistakes, clarify sentences, and refine the language for the intended audience or market.
In August 2009, Common Sense surveyed 27 corporations, two government offices, and two nongovernmental organizations that used human-assisted machine translation. Individual answers were not released publicly, but companies reported that HAMT doubled the translation output of what humans could do alone. The companies also reported that the hybrid method is up to 45% cheaper than using humans alone. Online tools such as Google Translate (GOOG) and Yahoo's (YHOO) Babel Fish are not yet accurate enough to do the job without humans, DePalma says.
Language translation is far from being mastered by humans, computers, or any mix of the two. Inherent obstacles, such as the speed of computers and the sophistication of software, are restraining the progress of automated translation, says Rayid Ghani, a senior researcher at Accenture Technology Labs (ACN) in Chicago. "More basic research and development is needed," he says. Much of that work will need to take place outside the U.S., he says, because of a lack of texts written in nonmajor languages for researchers to analyze.
The road to globalization, it seems, is paved in words.