Years ago when I was at Harvard University on a sabbatical midcareer break, I experimented for the first time with voice recognition software.

In those days, the technology had already advanced pretty well. That is to make the system work, you had to load the program on to your PC, then read aloud a long passage of text so that the machine could get used to your voice. It was all rather laborious, and not very accurate.

Things are different now. I have just downloaded onto my iPhone the cheap voice recognition software from Dragon.

No need to train the machine. You simply talk to the iPhone, and after a second or two the text appears. With an amazing accuracy. Once the text is written, you have the option of sending it by e-mail, by text or pasting it into Facebook or Twitter.

So I have recorded this blog posting, then e-mailed it to myself and put it up on the website. Very few mistakes. It all went far faster than I could possibly have typed it.

Quite brilliant. Rush off to the iTunes store and buy one for yourself. It’s hardly costing anything.

Update: I have indicated in bold where some errors were made. Pretty damn good for a first shot. 

Update:  It now has even managed Supercalifragilisticexpialidocious.

Update:  I reply to a reader’s thoughtful comments:

Thanks for the comment.

 

I am talking straight into my iPhone, so that you can see what exactly it produced. I have not changed any of the errors.

 

Where I remember to do it, I use punctuation. But it seems to manage to distinguish between different uses of the word eights [its].

 

Once upon a time I used to dictate quite frequently to my secretary. But alas I no longer have one!

 

I take the point about a favourable bias. I’ll have a go at talking about the problems in the Balkans to see what appears.

 

One of the most impressive things about this software is that it seems to work well however quickly you speak. I’m talking pretty fast at the moment, but the transcription still looks to be pretty good. You’re right of course about the difficulty with foreign names. I doubt whether many software packages will be able to cope with Bolton [Balkan] names such as you’d expect a rich [Izetbegovic].

 

In this case I have only used the dictating function a couple of times, and I’m already getting results as good as this. My impression is that the software has improved, although no doubt computer speed and other technical aspects are helping drive the improvement as well. In other words, the machine has had no chance to learn from lots of previous speech put in by me. I think this is dam [damn] good. My 10 year old daughter Ellie also has tried this a couple of times, with quite good results but not as good as mine-no doubt because she is giggling too much

 

Anyway, that’s enough for now. I’m going to press the button which pastes this text straight into an e-mail, and I’ll send it to myself and post it up as a reply to your comment. Best regards.