Disemvowelment Mode?

[Grant Rettke <grant@wisdomandwonder.com>]

** Disemvowelment Mode

When I do voice dictation I can keep up but it is unpleasant for me because I
have to type as fast as I can non-stop. I want to do a good job with the
transcription (few to no corrections required) and keep my fingers, wrists,
and mental health in tact.

Here are some things I researched towards that goal and my take on them:

- Use voice dictation software. Problem is that it can only handle perfect
  American English. It chokes on everything else.
- Use a faster keyboard layout. My setup is QWERTY and I am not going
to change that.
- Use the "jump back 15" seconds button to reprocess sections and correct what
  you had written. That is a painful experience that interrupts your dictation
  flow.
- Use a speed-writing approach. Very interesting. Not Shorthand because it is
  handwritten. Not Stenography because it is special hardware. Gregg Shorthand
  might work (http://gregg.angelfishy.net/djbfs.shtml) although it seems to
  require some handwritten parts. Problem with all of these is that they are
  only readable by someone with training in that approach. My dictation need a
  chance of someone else reading them, a good chance.
- Disemvowel is a portmanteau for vowel and disembowel—removing all of the
  vowels from a word for example
  "Tell me about that new xylophone they purchased for the country club"
  "Tll m bt tht nw xylphn thy prchsd fr th cntry clb". This is nice because it
  is simple and consistent. Your writing program can enforce this for you.
  Human thought is not required. Those are its strengths and weaknesses
  though—sometimes it is painful to read when it doesn't need to be. For
  example
  "Knock one more time then tell me yes or no"
  "Knck n mr tm thn tll m ys r n"
  is arguably just fine, or maybe not. My gt flng s tht thr s
  smthng bttr, vn f nly jst lttl bt.
- When I sit down and try to only use vowels I can do a better job then a
  program. I know that sometimes using a vowel makes total sense. For example
  "See me" becomes either a bad "S m" or OK "Se me" when I do it. The
  simplistic disemvowelment approach won't work here, and, that isn't
  surprising. It looks like there are some simple rules to make the process
  better. For example if the word is only two letters and starts with a vowel
  then it might be a good idea to leave it there. Another one is that if two
  words next to each other are converted to single letters next to each other
  then maybe they shouldn't be converted at all. There are things to consider
  here. I don't know what they are or what works well. But I think that this
  is part of the story.

Here is the mental space for the writer I'm describing here

- Groks the approach of removing vowels, but not always, because it needs to
  be readable.

The goal then is how to let the user just dictate and have to mode kind of
make things a little more correct after they are written. Sometimes the writer
can't do the mental translation fast enough and it would be easier to just
type the entire word. This is the situation for this mode: to standardize the
auto shorthand that keeps it true to how a human would do it.

What I want to do here is to make a mode that will help me a writer avoid
using vowels. The write would be typing along, making a point to avoid using
vowels. But sometimes determine that they are necessary, so it has to let the
writer type them in. The problem is that the writer won't always get it right
while trying to type fast. As the user is writing, the mode would look back at
the preview few words while it decides what to do with them.

It seems like there are a few ways to help the writer here:

1. If it sees a word with vowels in it, and if you disemvoweled a word that it
   only had 1 char remaining, then don't disemvovwel it.
2. If you see a word starts with a vowel, then leave it, but remove all
   the other vowels following.

Right now my ideas are not formed. They won't be until I try out the mode. I
haven been typing like this manually and it works fine, I'm just tired of
stopping and fixing the dictation because it doesn't look right (e.g. single
letter words). It is a break in the flow and it really hurts. Reading about
shorthand and also how people use abbreviations there are a lot of
opportunities to speed things up. For example for 'and' use '.' Instead of
'nd' and '/' to close a sentence. Yes definitely but that is out of scope here
for sake of keeping it simple.

From what I read this problem seems like auto-completion, backwards, for one
or two words, with no user interaction. I got this by reading about
auto completion functionality in one of the modes. So that is where I would
probably start.

What do you think? Where is a good place to start? Is there sometime like this
out there that I missed?

Thank you for spending your time on reading this, and thoughts and feedback.