CereProc: An Example of a Technology Finally Fulfilling Its Potential

Text-to-Speech is No Longer the Geeky, Awkward Stepchild of Sound

Glen Brizius
I can remember - back when the surface of the earth was still molten - I stumbled across a program on my computers hard drive called Dr. Sbaitso. It was a simple little program, which to be fair, most programs were in that era. You would type text into a box on the screen, and the computer would play the part of a psychologist / psychiatrist and reply back with some comment or question. Early, crude AI. But - and how cool is this - the computer didn't just think up a response and type it out on the screen. It said the response. Text to speech. How cool is that?

Sadly, Dr. Sbaitso was a little ahead of his time. (And a little out of his depth, caring for the psychological needs of an adolescent technology-obsessed boy). The technology just hasn't been there. Text to speech is one of those things that's always been sort of assumed (reference any number of sci-fi genre films and TV episodes), but the actual strides towards it have been fantastically difficult, and frustrating for those early-adopters who did at least try to address this market. To be blunt, early text-to-speech software was embarrassing bad.

The core problem is that speech is just too darn complex. Even in a relatively non-tonal language like English (which doesn't rely on the pitch of the speakers voice to be precise - up on that syllable, down on that one, etc) there are still multiple ways one can say a word, and that word will then have any number of inferred meanings based upon that inflection. It might be dripping with sarcasm. It could be a statement posed as a question, accomplished by an upwards tilt of the pitch towards the end of the sentence. If a picture is worth a thousand words, a spoken word is worth at least a hundred written ones - just one of the reasons I'm so anti-IM technology, but that's a rant for a different time.

Here's the thing: in the past five years, a company has (relatively quietly) sprung up and is making serious headway towards fulfilling the dreams of Dr. Sbaitsos early patients. The company is called CereProc. It's a Scottish company, based in Edinburgh, although they have sales offices in other locations throughout the United Kingdom. The name CereProc is short for 'Cerebral Processing', and the company was founded in late 2005 by Matthew Aylett and Nick Wright of the Edinburgh-Stanford Link speech research fund. The pair of them (much like myself) were dismayed by the lack of progress being made towards speech technology. They obtained grant funding from Scotlands government and went to work, hiring the very best in the speech synthesis field to come to Scotland and make serious progress towards the goal of having a computer analyze text and produce it as a voice through the speakers.

Well, that grant money paid off big-time. For the past five years, CereProc has experienced double and even triple digit year-on-year growth, which is a boom by my definition. They've added a lot of products to their line of unique applications. The companies core product, CereVoice, is available on any platform. Mobile phones, embedded chips, desktops, servers, headsets, you name it. Aspiring speech application developers can use the CereVoice Software Development Kit (SDK), which contains the key elements to produce new applications, to incorporate CereVoice text-to-speech technology into their products. This licensing is free for academic institutions doing research, which will seriously fuel the honing and sharpening / general improvement of this technology. Academic institutions love free licensing, and the company wins as well by getting feedback from such a diverse user base.

The litmus test - does it work? Or is it, like so many of its predecessors, simply embarrassing? Decide for yourself:

http://www.cereproc.com/support/live_demo

So what's the bottom line here? Put simply, CereProc is going to continue to operate much as it has these past five years - in the background. It will probably never be a household name, but it's likely that many households will experience its technology as companies begin to incorporate CereProc into their products. It's the extra layer of polish on top of that brand new car that you admire - you can't quite put your finger on it, but it adds that certain something. Applications will become more intuitive, and humans will take one step closer to fully embracing the machine as one of their senses becomes more entwined with the technology that surrounds us.

References:

1. http://techstartups.blogspot.com/2010/03/cereproc-text-to-speech-techstartup.html

2. http://www.cereproc.com/
3. http://en.wikipedia.org/wiki/Dr._Sbaitso

DISCLOSURE OF MATERIAL CONNECTION:
The Contributor has no connection to nor was paid by the brand or product described in this content.

Published by Glen Brizius

My name is Glen Brizius, and I'm a 36 year old resident of the USA. During my life I've dedicated my life to science - particularly, the study of chemistry and polymers, or plastics, as most people know them...   View profile

To comment, please sign in to your Yahoo! account, or sign up for a new account.