Google claims that prototype real-time speech translation systems are performing with "close to 100% accuracy" in laboratory conditions, but has given no indication as to when a finished product will become a reality.
Google Translate, the web and mobile tool for converting text
from one language to another, is already indispensable for most business travelers, expats and holidaymakers. Being able to type in a foreign text and get an immediate translation has saved countless people from accidentally ordering horse in a French restaurant or from using the wrong public toilets. However it's only as effective as the user's keyboard skills and when it comes to symbol- rather than character-based languages, unless the translation in question is for a website and copy and paste is possible, well, you're on your own.
To help alleviate this problem, Google has already added a ‘free-hand' feature. Launched on July 24, it lets users move a finger on the screen or a mouse on the desktop in an attempt to ‘draw' the characters in question, and while that's a good start, in its next trick it hopes to remove the keyboard altogether.
The company is developing a real-time translation tool that can do for voice what its current systems can do for text. In an interview with The Times, Android product management vice president Hugo Barra revealed that the current prototypes are achieving "close to 100% accuracy," and that with some language pairings, the results are "near perfect," meaning that in laboratory conditions at least (i.e.,: without background noises and over a perfect internet connection), two people speaking two completely different languages can communicate via the system.
Earlier this month, Microsoft Asia researchers demonstrated a system that uses the Kinect sensor that can translate sign language in real time and also turn text into visual sign language via an avatar. And while this breakthrough will no doubt have a huge positive impact in making not only the web but a number of public spaces -- say transit stations or hotel lobbies -- more user friendly for those that are hard of hearing, there are no end of applications for a mobile, real-time speech recognition and translation tool.