Wednesday, November 19, 2008

My Hovercraft is Clogged with Eels

Mr. Spanky is happy to report that he now has a Facebook page. I know that many of you kidz are thinking, "This is *so* embarrassing, it's like he just doesn't *get* his target audience. He should have *started* on Facebook." Once again you underestimate the sheer brilliance that Mr Spanky can barely contain in his shiny metallic little cranium.

I can now post to Blogger (i.e. my blog) and it automatically emails out to those of you that are on my mailing list. Simultaneously it posts to my Facebook page. This is just unprecedented genius. You can consume Phun Kidz in any form that you like! And I can be lazy or as adults like to say, efficient. If you are ever in the position where you want to know how to do what I did then send me an email. I will make you pay dearly. In the mean time go to my Facebook page and add me as your friend!

But enough jawing about my amazing skills. Let's get on to the interesting post of today. For those of you that are lucky enough to have an iPhone then you should know that the absolute killer app has arrived. And get this: it's free. As in no cost from the gearheads at Google. It allows you to perform a Google search by voice from your phone. I tried the voice recognition on a number of phrases: "Mediterranean food in Mountain View", "Starbucks in Willow Glen", and "She turned me into a newt". Amazingly, every single one of these phrases was turned from the croaking voice of Mr Spanky into flawless text for Googley consumption. More esoteric phrases like "My hovercraft is clogged with eels" caused it consternation, however. The big question, kidz, is how in the name of Gibil's flaming nethers does this work?

As you speak into your phone, your voice is transmitted to a hundred thousand monkeys on a secret Google monkey farm in some desolate spot in the world like Meat Cove, NS. Each of them wears a headset that shouts your voice into their pea size brains. They are trained to punch a phonetic keyboard with all the basic English sound clusters in response to bananas that hang temptingly from the ceiling. The results are averaged and sent back to Google in order to perform the search. Or something like that.

In reality your voice is transmitted to the Google servers and they use some Text-To-Speech software to break it down and guess what you said. The technology has been around for quite some time and has improved steadily over the years. Google has no doubt scaled it to handle the volume of translations needed. I suspect that they've also allowed the software to learn from mistakes as well. If the speech is not translated exactly the way you want then there is a little drop down with other possible variations that you can select. I bet that once you select a variation then the software "learns" where it made a mistake. See neural nets for what I mean by "learning".

Conceivably, this technology could go both ways. You might eventually be able to utter into your iPhone in English and have your speech broken down to text and then translated to Hindi and spoken using Text-To-Speech software. Just like the Pomegranate. I took the liberty of running the English phrase "Mr. Spanky couldn't see more than that because of the fog and rain." through Google translation services into Hindi and then back to English: "Mr than that Spanky because of fog and rain could not see more." Clearly we still have some way to go.

1 comment:

bubba said...

says on facebook that you are married. i knew you'd eventually hooked up with darla, you dawg.