Wednesday, February 10, 2010

It’s all getting a bit Star Trek - Google is working on voice-to-voice translation

ust the other night, as a friend of mine watched YouTube videos over his Wii on his plasma-screen TV, he turned to me and said “sometimes I feel like I’m living in the future.” And I know what he means. Reports have surfaced that Google is working on voice-to-voice translation software, that would take what you say in your language and translate it into another.

What’s the story?

According to the Times Online, Google is working on combining two existing services. It already has Google Translate, and it has a voice recognition system that you’ve probably used if you have Google Mobile on your phone. The idea is that the voice recognition software will transcribe what you’re saying into text, and feed that into the translator. You can then select the output language, and the service will “speak” the translated text.

What’s the ideal?

I remember when I used to watch Star Trek as a kid, the ships computer would just translate whatever people said to one another. It’s a common problem in Sci-Fi shows - the fact that people all seem to speak the same language. Often its explained away with a quick “the computers are doing it”.

It would be incredible if that were no longer just fiction. The ideal application of this service, for me, is on mobile devices. Imagine you could walk into a restaurant in a foreign country you’ve never visited before, tell your phone what you want to say, and have it relay it to the waiter. And then you could use the phone to translate in reverse.

What’s the reality?

Even modern text-to-text translators are pretty patchy. And I’ve never seen a voice-to-text transcription service that really, REALLY works. Some have gotten close, but they almost always end up pretty shabby. And if you’re talking about using it for lots of different languages… even different accents across the English language can cause voice recognition hideous problems. Imagine trying to deal with other dialects!

I can only imagine being on the receiving end of this service - a foreigner walks up to me, speaks into their mobile, and it proceeds to spew a jumble of mismatched words that barely make sense.

Hey, maybe I’m wrong and it will be awesome. But probably not.