The translation industry has been debating the human versus machine subject for many years. We have heard repeated claims that machine translation was the next big thing. But to date, the technology has never quite arrived. Software has improved but language changes regularly and can have multiple meanings.
The Obama administration has now challenged companies to develop software that will produce accurate translation of the major languages of the world. This comes with the offer of cash for research which could ignite new technologies and lessen the language barriers that may get in the way of international business. The actual amount of money that will go directly to translation efforts is not yet clear, the industry has grown markedly for many years and with the recession, there are plenty of linguists looking for work.
In a global marketplace, companies need to advertise their products in many languages, notably on their websites where research shows customers feel companies that speak to them in their own languages are more credible. A 2006 Common Sense report shows that 52% of consumers surveyed said they would only buy something from a website in their own language, whilst 64% said they would pay more for a product if they could get information about it which they could read. Indeed, businesses often find savings by using the languages of countries to which they are marketing.
The most advanced translation providers use human-assisted machine translation (HAMT), where text is fed into a computer programme that does the first round of word and sentence conversion using statistics, language rules or matching with past translations. This covers roughly 90% of the work and then a human corrects mistakes, clarifies sentences and refines the language for the intended marketplace.
It is generally agreed that language translation is a long way from being mastered by humans, computers or any mix of the two. The progress of automated translation is also impeded by obstacles such as the speed of computers and the sophistication of software.
This debate will, it seems, continue.