A universal Google translator will have linguistic rivers to cross


This article was originally on a blog post platform and may be missing photos, graphics or links. See About archive blog posts.

Google is planning to break the language barrier and deliver a phone within the next few years that will translate languages on-the-fly, a company representative told the U.K.’s Times Online in a recent interview.

‘We think speech-to-speech translation should be possible and work reasonably well in a few years’ time,’ Franz Och, Google’s head of translation services told the outlet.


The idea behind Google’s plan is simple: when a user speaking one language says something into a phone, Google’s translator software, which would be running on the recipient’s phone, would interpret what the person said, translate it into the recipient’s language, and recite it back in that second language. In order to limit conversational pauses, the company apparently plans to translate as phrases are spoken, rather that waiting until each sentence is completed.

Google hopes to make the application available ‘in a couple years.’

Difficulty awaits

In practice, achieving real-time digital translation will be no easy feat.

Voice-recognition technology has traditionally struggled to work quickly and accurately. Many services built in to current mobile devices require rigid commands to function. For example, in order to initiate a call using my Bluetooth headset, I’m required to clearly say ‘call’ followed by the person’s name. I’ve got to speak like a robot and even when I do, the software sometimes fails to understand my commands.

Even Google Voice, the search giant’s telephone service, has difficulty translating speech to text. Whenever a caller leaves a voicemail on a Google Voice user’s answering service, Google delivers a transcript of the call to the user’s voice inbox, which is similar to an e-mail inbox. Many users have found that those transcripts are...

...rife with errors.

The trouble with complexity


As wobbly as voice recognition services are, their complexity pales in comparison to what Google is attempting to achieve with its phone-based translation software. Google’s application will need to accurately determine what a person is saying in any given language and then translate all those words into a new language. How Google will ensure accuracy in languages that have identical sounding words with different definitions is a major question mark. The same can be said for slang, dialects and other language quirks that will only complicate the translation process.

Once Google’s software translates the caller’s words, it will then need to convert them into voice to relay the message to the person on the other end of the line. When that happens, Google’s software will lose voice inflection and hidden meaning. Nuances of speech like sarcasm, irritation and excitement will be in danger of being lost in translation.

-- Don Reisinger