Vestec: Shut up, the angels are talking!
I’ve never been a big fan of speech recognition as a primary interface. No matter how good the technology, controlling machines by talking to them is awkward and time-consuming. You need a thousand words to describe a picture.
I guess some angel investors wouldn’t agree with me – Waterloo-based Voice Enabling Systems Technology (that’s a mouthful) or Vestec – just closed a round of financing. Vestec claims to be different than past voice UIs by not requiring any user/system training, and being able to understand natural phrases in many different languages. I’ll believe it when I hear it!
A more useful and reliable “voice” interface is text to speech translation. There are many applications for this: reading out warnings and notifications, accessing documents or messages remotely… For a demo, try out Bruce Tsuji’s SpokenText.
I’d agree that voice commands are a pain – and make the design of Speech to Text systems difficult – because in addition to decoding speech it always has to check if each word is a command.
This slows down the feedback to the speaker introducing delays between spoken words and user correction if the system is wrong. The result is that learning the users speech is slower than it could be.
Still having just started using a speech to text system for blog posts it’s a great way to write.