Thursday, May 27, 2010

You know my Blogger profile icon?



That's a picture of my dear old Gemini2, which I bought back in 2006, while I was still in steno school. My friend Mel (who I met through my friend Sumana) is here from Boston, and I'm lending her my Gemini2 so that she can figure out whether learning steno might be helpful for her RSI issues. We're having a bit of trouble hammering out the Python2/Python3/pyserial/Tkinter peculiarities on her system, but I hope we'll get the hang of it soon. I'm really excited to have a professional Free/Open Source Software expert in on the project!

2 comments:

John said...

Hi Mirabai Knight,

John McKown here, the Chordite guy. Every now and then I google Chordite to see any new references and thus found your blog. Here's an unsolicited brain dump.

May I say what you will certainly already know: you write very well --- commercial quality one might say. I'm curious do you do it in steno :-)

I think a person who needs to _create_ text faster than even a slow typist is either very smart or a pretty sloppy writer. Being neither I typically pound out a few words and then replace them. Even at 35wpm I can type faster than I can think what to say or, more exactly, faster than I can express things to my own satisfaction.

Of course authoring and transcribing have different speed requirements (or should have). With steno you can type as fast as people can rattle on, however thoughtlessly. Except for court reporting it's hard for me to see that as a good thing. :-)

I would beg to differ on chording. I find it distinctly faster than a qwerty, probably because the finger:key correspondence is 1:1 and fixed. But perhaps that's just me.

I see you realize that steno will never be widespread as a general purpose computer interface. Most people are far too lazy to learn it. But also, why should they. When would they need to type that fast. Better to wait for better speech recognition.

On the other hand what I'm doing here is extrapolating from myself, saying well, I couldn't use it therefore no one can. And that's probably not completely right

Mirabai Knight said...

Hi, John. I've been lurking on the Chordite mailing list for years, and while I don't think it's a satisfactory input system for my purposes, I'm very glad it's out there. It's an inspiring project. Thanks for the kind words about the blog.

The issue with composing text in steno, as I addressed in part two of my "What is Steno Good For?" series (http://plover.stenoknight.com/2010/04/writing-and-coding-with-steno.html), is less about speed than it is about fluency. Even if you can only think of the words you want to write at around 40 words per minute, in qwerty you have to spend most of that time pounding each individual letter; it's a lot of extra stuff to be doing with your brain when you're trying to think of what to say, and it disrupts the smooth flow of composition. In steno, you make one stroke per word, and if you want to pause in between those words to think of what to write, you can do that without having to tip-tap-tip-tap every letter. I'm not a particularly quick writer, but if you read the article I linked above, I found myself far less afflicted by writer's block when I composed a book in steno than I did when I tried doing it in qwerty, partly because I was able to get into a smooth, comfortable flow state so easily, without having my fingers constantly in agitated motion.

When you say "better to wait for better speech recognition", well, that's a whole story in itself, but I fall firmly on the side of believing that speech recognition will always have enough errors that any person dictating to a computer will either have to be constantly vigilant for mistakes or to edit the transcript extensively afterward to get rid of errors. I'm sure that's fine for some people, but I like more determinism in my speech input techniques.

Here's an article discussing why speech recognition technology is unlikely to get significantly better than it is already. It will continue to improve somewhat, but the improvements will be incremental and slight, and approach an ideal state asymptotically that is still far beneath the necessary accuracy threshold of many people:

http://robertfortner.posterous.com/the-unrecognized-death-of-speech-recognition