OK computer

I tried out Windows 7's own speech recognition software for the first time yesterday, and I'm nearly, almost, amazed. It'll take a little while to work out what I really think about it. It is, overall, probably faster than typing. But typing is a very different process. I'm reminded of Julia Cameron saying that it's all about getting stuff down, not thinking stuff up - the direction's important. Speaking feels like thinking something up. I guess that could change with practice, but the error-rate of typing doesn't interfere with the flow in quite the same way. (Don't interrupt the flow, man... )

I just did a quick one-off test. The paragraph below is the opener for an awesome 1979 geography textbook, 'people, pattern, process', by Keith Chapman, from the days when human geography was just starting to wonder what all this critical theory business might be about. The one I include here was spoken, and I've left in all the errors it made, with corrections in square brackets. Despite my best efforts at newscaster-speak, I still mumble, and perhaps the mic isn't great (just a bog-standard headset) - but this is still pretty impressive given that I didn't need to lift a finger. So: spoken averaged 83 words per minute. I read that people can speak at 120, but that would be a fair old rush. My typing attempt averaged 40 wpm: I'm quick in short bursts but pretty error-prone so I spend a lot of time correcting, but the end product is at least accurate. That 83wpm doesn't include going back through and fixing things.

Even so, pretty impressive. Somehow, even 'Star-trekkers' presented no problem. The program also goes through and indexes your documents, so it's been managing very well at replicating both my own academic language and idiosyncracies (including 'workinz'... perhaps not ideal.) It also allows for easy correction, which it learns from, but that does of course slow things down.

When I have the money, paid speech software might be worth a go if it would improve on this already pretty smart program. But we'll see: perhaps, as I say, it doesn't come down to speed. Words through fingers are different to words aloud. Anyway, here's the speech program's attempt at interpreting my reading, done in 1 min 53 seconds as opposed to about four minutes of hapless typing:

Good science fiction should maintain a credible link between reality and imagination. A productive think of it as [should be: theme for writers of] science fiction has been man's ability to jump be on [beyond] the barriers imposed by the dimensions which define his existence - space and time. Thus HG Wells time traveller could project himself but [both] forward and backward in time. The 'transport are' [transporter] of the starship enterprise enables Captain Kirk [it knows that's a name...?] and his crew to travel through space instantaneously, although it has been known to permit simultaneous movement in both dimensions! [star trek episode in-joke in first para, wow!] Geography may appear to have little in common with such whorls [worlds. I prefer Window's word] of the imagination, but his [its] position as an academic discipline is related to its explicit concern with spatial relationships of objects and events at the surface of the earth. The universal availability of the kind of technology at the disposal of the Star-trekkers would transform these relationships by effectively nullifying the role of distance as an obstacle to movement between one place and another.