Future Processing on advancements in speech recognition

Do we still need to learn foreign languages?

date: 22 September 2016
reading time: 2 min

How many times have you seen someone trying to have a decent conversation with their phones or tablets? We’re guessing many. And yes, it really is painful to watch sometimes. But can technology make it better by improving speech recognition software?

Deep learning needed a lot of time to bring us speech recognition programmes we’ll be satisfied with. Whereas in 1980 speech recognition might have been something unheard-of, today it is expected to represent some level of brilliance.

And here we are, experiencing it to the fullest.

As it turns out, there is an innovative system able not only to transcribe foreign languages but also do it faster than humans can type. How? Thanks to Chinese search engine Baidu, which introduced a program called Deep Speech 2, able to do just that.

It has also been discovered that the system produces 20.4% fewer errors than people typing in English and 63.8% fewer in Mandarin!

Quite astonishing, isn’t it?

Baidu researchers conducted an experiment with 32 volunteers (19-32 years old) where some used QWERTY phone keyboards with all their features like auto-correction and suggestions and the others spoke random phrases like ‘Have a good weekend’.

The system still needs improvements and Baidu doesn’t plan to make it available to the public for now. It will be integrated into Baidu apps, though.

Even though Baidu wants to keep the software secret to themselves. The creation of such system is possible and it is only a matter of time before some other company will recreate their software.

So, can we stop learning languages and let translators catch their breaths? Not really.

Liked the blog post?

Please give us feedback.


This website stores cookies on your computer. These cookies are used to improve our website and provide more personalized services to you, both on this website and through other media. To find out more about the cookies we use, see our Privacy policy.