Google Challenges Voice Translation
"Google hopes to bring instant voice translation for phone conversations."
Published: 8th February 2010 | Source: The Sunday Times |
Over the years, Google has been hard at work expanding the scope of its services. One service that has gotten a strong service is Google Translate. What started as a simple, six language translator has since grown to encompass a total of 52 languages. Now it seems Google is looking to push it to the next level with voice translation.
To do this, Google will be pairing their translation service with their voice recognition system currently used to allow phone voice commands. Speech will be taken in, processed/translated almost immediately, and then fed through a typical synthetic voicer to the receiver. The aim is to allow foreign users to have phone conversations with each other speaking in their own native language.
“We think speech-to-speech translation should be possible and work reasonably well in a few years’ time,” said Franz Och, Google’s head of translation services.
Two main issues with speech-to-speech translation come to mind. The first involves the actual translation. While Google's translator is certainly more likely to be correct than alternatives with its use of the Google database and statistics, it still has difficulties with nuances of languages, particularly with informal speech and slang. A prime example of this is shown by the site Translation Party, which takes a phrase and translates it back and forth between English and Japanese until a state of "equilibrium" is achieved.
The other concern, brought up by experts like David Crystal (honorary linguistics professor at Bangor University), involves the voice recognition system's ability to deal with the vast number of accents heard around the world. Crystal feels that there are just too many out there for any current system to cope with.
“Maybe Google will be able to get there faster than everyone else, but I think it’s unlikely we’ll have a speech device in the next few years that could handle high-speed Glaswegian slang," says David Crystal.
Och responds to these concerns, confident in Google's software and its ability to adapt.
“The more data we input, the better the quality,” said Och. There is no shortage of help. “There are a lot of language enthusiasts out there,” he said.
“Everyone has a different voice, accent and pitch,” said Och. “But recognition should be effective with mobile phones because by nature they are personal to you. The phone should get a feel for your voice from past voice search queries, for example.”