During a recent Microsoft Research event in Tianjin, China, Microsoft has demonstrated that it can almost instantly translate spoken English into spoken Chinese.
The breakthrough in this does not lie as much in the capability to translate into another language, but more so in understanding what a person actually says. Computer scientists have been involved in the development of this breakthrough for the past 60 years.
Instant translation of speech is particularly difficult because of the struggle to capture the range of human vocal ability, the difference between human’s voices, the time it takes to convert data, and the lack of time to correct inaccuracies.
Until recently though, even the best speech systems still had word error rates of 20-25% on arbitrary speech. The word error rate has now been reduced to 30%.
Needless to say, the system has no real impact on human translation yet! An error is still likely both in the English text as well as in the Chinese translation, but is certainly a technology to look out for!
Please follow the following link, to listen to this speech: Instant Speech Translation