Translating research into business solutions

Google Translate is making strides in helping businesses function in a world where geographical boundaries are becoming increasingly fuzzy, says Ashish Venugopal, research scientist at Google.

"Content on the Web is dramatically changing from being English-centric to very multilingual," he said, explaining why Google Translate plays an important role on the online platform.

Google's research shows that while Arabic content constitutes only 3.3 per cent of content online, it has grown 20-fold over the last decade.

Chinese is an equally striking example: 22.6 per cent of content on the Web is now in Chinese - a dozen times more than a decade ago.

"When the Web is changing like this and when there's content being produced in all languages, it becomes even more important that we can all understand the content that's being produced," he said.

The service uses a statistical machine translation system, as opposed to a rules-based system of machine translation built on grammatical and syntactical rules.

The statistical system relies on information published on the Web in multiple languages, from blogs, news sites and government sites, among various sources.

Based on that large body of data, the system matches word and phrase patterns that occur between languages to produce a translation.

"When you give us a new document, we find all the examples of the words in that document that have been translated by humans when they created those documents, and try to match it up and produce a foreign translation for you," explained Mr Venugopal.

He also attributed the rapid growth of available languages in Google Translate to the statistical system engine running behind the service.

The service began with translation between English, German and Spanish in 2006 but can now translate between 64 languages.

He said: "It's trying to learn how to generally do translations for all pairs of languages, as opposed to a company sitting down and saying 'I just want to do translation for this pair of languages'."

The statistical system is also very robust, the Google official says.

Human speech changes over time, with slang words like LOL becoming accepted vocabulary.

But because the system "learns from what people use on the Web, it simply adapts over time to the way people speak", he said.

Google Translate "works very well on average", but that "doesn't mean it won't make mistakes".

While it can generally give users a "clear perspective" of what they input into the service, grammatical errors and such might still occur.

This is because translation is a complex process that comprises many small steps - for example, changing the word order.

Mr Venugopal said: "All of these different operations are required to make a translation come out fluent by preserving the meaning and sounding smooth, but at any point of this process, some of these can go wrong."

Which is why Google Translate allows users to choose alternative translations for specific words and so contribute feedback to the system.

"Right now, our goal in machine translation is not to provide you the perfect translation; it's to provide you the meaning that used to exist in the original language.

That's the goal of machine translation and how we intend it to be used."

Since Google Translate is built upon human translation, professional translators and interpreters worried that the advent of such online translation services will slowly erode their market value can put their minds at ease.

And in any case, that is not Google's intention.

"Our goal is to make machine translation available to break down a language barrier. That does not preclude the need for human translation and human interpreters to be involved in the process."

In fact, Mr Venugopal believes that Google Translate can complement the services of such professionals.

For instance, there was a company that previously had to engage a human translator to interact with another company in China, even for the simplest of transactions.

"Business was effectively slowed down . . . But after they switched to Google Translate, they were able to take advantage of the human to provide the highest value," he said.

Simple deals were undertaken with the online service, but when negotiations became more complex, the company engaged the translator - who also benefited in that he "was able to be involved in many more transactions that he'd otherwise have been able to".

Mr Venugopal added: "On a higher level, if a prime minister or a politician gives a speech, our translation will be effective at telling you what he said, but not necessarily how he said it. If you want to convey that nuance, the tone and underlying feeling of a speech, you're going to have to bring in the human translator."

Google also released a Translator Toolkit in June that aims to make translators more efficient at their work.

The toolkit gives translators a quick translation of uploaded documents, which the translators can then edit.

Businesses can also leverage Google Translate to expand their consumer base beyond local shores.

With Google Translate, businesses can read the latest news about their sector when it is published in a foreign tongue.

They can also make their websites available in more than 60 languages with the free Website Translator plug-in, which even allows site administrators to customise and improve translations for their sites.

"The world is becoming a more and more connected place. Companies should be looking at the whole world as their consumer base, as opposed to a language-specific group of people," Mr Venugopal said.

This article was first published in The Business Times.