vGoogle - voice crossing over into the net
13 years ago, in a paper featuring a "spherical support analyst", I talked about the information we all lost in p2p voice conversation, all the bits we'd need to store, etc. It was astonishing how much information we lose on a daily basis.
It turned out that the value of the stored conversations, particularly by those engaged in problem solving - support people, lawyers, doctors, engineers particularly - was calculable.
The specific case I took on was in tech support: On average - We had 70 support engineers taking 20 calls each per week, spending over five hours on the phone per day, solving roughly 40% of the problems with an internal document, and writing an average of 1 new internal document per month... that meant that ~800 hours a week worth of problem solving - the hard stuff - was going uncaptured, unrecorded, undocumented. A tremendous amount of information lossage! And it was/has been accepted as normal - that calls "are only recorded for quality control purposes" - not for feeding back into a searchable, indexed database of problems and solutions...
(I note that the documentation problem has been solved today in another way - all the free information accessed via google written by smart netizens far outweighs what any company can do nowadays with a proprietary database) - but the issue remains - thousands of hours of useful conversations are being lost, every day, in every business. Even if 90% is crap, that last 10% would be a corporate IQ enhancer on par with google itself.
I've been recording conversations, on and off, ever since I wrote that. I've had some fun with the idea - (why have a conversation with someone in your house when you can have it just inside the taj mahal, on a light rainy day, with some elegant music playing in the backgroun?), but was never inspired to try productizing the various technologies.
A couple years ago I realized that it was now possible to actually record and index *everything* I said over the phone, and started working on making
asterisk do that.
I tried to explain it to people then, calling things like "vgoogle" and "voogle", trying to explain the hidden value in making voice recorded - and searchable - and met, generally with a blank look - or the over-enthusiastic leap to thinking we could go direct voice-to-text (The principal barrier to that is that speaker independent voice recognition still hasn't materialized - and voice recognition in a large domain still requires very powerful computers)
But... if you could have software that could just recognise a few key words and phrases accurately - and index their places in the call - or if you had a service that could transcribe, using human resources, cheaply - the value of the recording goes up. An hour of transcription costs about 60 bucks in India...
Last year I came up with a few markets where I thought the technology would be useful, and settled on conference transcription as being the first market that made real sense, and called the
prototype service "transconf". But I ran out of money, and time, and I gave up, rationalizing that "It's not something I have to do". "It's an idea that obvious once you think about it long enough", and I knew that some company out there would get the idea, eventually, and make a killing. I thought of Heinlein, and the water bed...
Then
podcasting happened, and we started seeing
a need for devices that had this sort of indexing...
And about 6 months ago,
conference providers started providing recording services... and I felt that it would only be a few more months before some service started to get the idea that recording and indexing conversations would be a big market.
Today,
Now google is getting into voip. The techniques and utility of bridging the voice world into the data world are going to become obvious, real soon now, and it's going to become a multi-billion dollar market.
I'm a little grumpy. I wish I'd been smart enough to patent something in this field all those years ago...
Anybody wanna hire a visionary, cheap?