Re: Re: Info about DVORAK Keyboard Design

Hi all,

Here I include my response to kattdood's email, since he had emailed me privately about this.


> >  
> >   I use this site as the reference for frequencies:
> >   http://deafandblind.com/word_frequency.htm
> >   Many interesting lists there. Look at the top twenty words most 
> > used in written English: not a single U.
>   
That is fine, but looking at the top twenty words is an incomplete
analysis. They also don't say how they arrived at their statistics. I
used the following program to analyze samples of English text from
Project Gutenberg:

http://millikeys.sourceforge.net/freqanalysis.html

As you can see, the developer who wrote the program analyzed more than
1,400 works of  fiction. He got OU as the 22nd most common digram. IO
does not even appear among the most common digrams, since he truncated
his list.

In my own analysis of 8 english works, I got OU as the 25th most common
digram, and IO as the 92nd most common digram.

I used the following 8 works:

A Connecticut Yankee in King Arthur's Court by Mark Twain
Charles Darwin - On the Origin of Species By Means of Natural Selection,
or, the Preservation of Favoured Races in the Struggle for Life
Dostoevsky - Crime and Punishment
Life on the Mississippi by Mark Twain
The Adventures of Huckleberry Finn by Mark Twain
The Antichrist
The Unabomber's Manifesto
The UNIX-HATERS Handbook

I got most of them from Project Gutenberg:
http://www.gutenberg.org/wiki/Main_Page

I encourage you to download the program, get some English text files,
and do your own analysis. I used the program to design my own
international Dvorak version. I often type in Spanish and need accented
characters. I also wanted the numbers laid out as in Dvorak's true
layout, and I hated where ANSI put the question mark.

I also modified this java program to test my own layouts:

http://www.acm.vt.edu/~jmaxwell/dvorak/comparePage.html



> >  The index finger travels around a lot already. Putting the I 
> > there means it stays home more often, so to speak, because I is 
> > used more than U. If U stays on the home row, it's still quite 
> > easy to type its vowel-vowel digrams.
>   
What you are not considering is that not all finger moves off the home
keys are equally costly. A move off the home keys when the prior or
following keys are typed with the same hand is more costly than a move
off the home keys when the prior/following keys are typed with the
opposite hand. The reason is that when the digrams are typed with
opposite hands, the finger on the opposite hand can position itself in
advance, while the other one types. With Dvorak, the TI combo isn't so
expensive, because your index finger can move to the I, while the other
hand types the T. In contrast, OI is more expensive in Dvorak because it
is typed one-handed and you cannot take advantage of alternation.

So, to swap the U and the I would make the common OU, AU, JU, and QU
much more expensive. Swapping I and U would make OI easier to type, but
OU would become harder, while the benefit for typing digrams would TI
would be insignificant because those are typed with opposite hands. And
since OU is much more common that IO, a good case can be made that U is
in the right place in the Dvorak layout.


> > Statistically, I believe it is slightly 
> > more efficient than Dvorak. This is a testament to how well-
> > designed the Dvorak layout is.
>   
I am familiar with the Klausler layout. It is good, but there is an
important factor that it does not take into consideration, which Dvorak
does: inboard strike flow. I am not sure if you are familiar with this
concept, but I'll explain just in case. The idea is that it is easier to
type starting with the pinky finger and then flowing inwards towards the
index finger. You can try the concept yourself by rolling your own
fingers on a table. You'll find that it's much easier to do when you
start tapping with the pinky, followed by the ring finger, then the
middle finger, and ending with the index finger, than to do it in the
opposite direction.

For example, the Dvorak WH, CH, SH, TH, NT, ST all are typed from the
outside fingers towards the inside fingers. 

Klausler's layout looks as follows:

k , u y p  w l m f c   Best evolved layout
o a e i d  r n t h s   28281895
q . ' ; z  x v g b j

It is clear that the second most common digram in English, TH, is typed
outwards. Other very common digrams, like NT and NG, are also typed
outwards. WH is not only typed outwards, but is also very awkward.

Just looking at the layout, it is obvious that Klausler's algorithm is
not taking inboard strike flow into account. This may be why, when he
used real data, he could not produce a layout better than Dvorak. The
only significant benefit I see to Klausler's layout is that it relieves
the right pinky.

Here is a site where you can compare inboard strike flow in Qwerty,
Colemak and Dvorak:

http://www.siteuri.ro/dvorak/

Using the first 5 paragraphs of the Unabomber's Manifesto I get the
following results for the outboard strike flow statistic:

QWERTY 12.5 % 
Dvorak 6.6 % 
Colemak 11.5 %

So Dvorak has about half as many digrams typed outwards as do Qwerty or
Colemak. The site does not let us compare Klausler's layout, but just by
looking at it, I think it would fare as poorly as Qwerty, and possibly
much worse.


> >   A page about alternative keyboard layouts that compares the 
> > various well-known layouts (and some experimental ones):
> >   http://sagez.net/amuseum/layout.php
>   
Yes, I use that program myself to test my own layouts. However, that
program has many limitations. Note that it does not measure
inboard/outboard strike flow, and so does not do justice to Dvorak. It
also does not measure the incidence of row-jumping. Even more
importantly, it only looks at statistics for individual keys, and almost
completely ignores digrams. If you think about it, people actually type
digrams and trigrams, not individual keys. The program also neglects to
measure the statistics for the home keys (as opposed to the home row).

Thanks for the link though. It contains a few oddball layouts that I had
not seen!

Best Greetings.



kattdood wrote:
>
> Actually, EA and IO are the most common vowel-vowel pair. OU
> is third. also realize that TIO and ION are very popular
> trigrams, and this is, at least for myself, a hindrance to type on
> the current Dvorak layout.
>
> The letter I being so much more common than U means that
> it should be much easier to type. Not just the first-person I,
> but common two letter words begin with I, is, it, in, if. It's
> best not to stretch your fingers to repeat the same key, such as
> Roman numerals.
>
> I use this site as the reference for frequencies:
> http://deafandblind.com/word_frequency.htm
> <http://deafandblind.com/word_frequency.htm>
> Many interesting lists there. Look at the top twenty words
> most used in written English: not a single U.
>
> The index finger travels around a lot already. Putting the I
> there means it stays home more often, so to speak, because I is
> used more than U. If U stays on the home row, it's still quite
> easy to type its vowel-vowel digrams.
>
> If you're really interested in a Dvorak-like layout, checkout
> Klausler Evolved, which was the result of an Evolution Algorithm
> program. The U is on the top middle finger, so vowel-vowel
> digrams are still pretty easy. Statistically, I believe it is slightly
> more efficient than Dvorak. This is a testament to how well-
> designed the Dvorak layout is.
> http://klausler.com/evolved.html <http://klausler.com/evolved.html>
> Pic of the winning layout in PDF
> http://klausler.com/evolved.pdf <http://klausler.com/evolved.pdf>
>
> A page about alternative keyboard layouts that compares the
> various well-known layouts (and some experimental ones):
> http://sagez.net/amuseum/layout.php <http://sagez.net/amuseum/layout.php>
>
> --- In [email protected]
> <mailto:altkeyboards%40yahoogroups.com>, Don <egr330@...>
> wrote:
> >
> > Hi,
> >
> > Dvorak didn't just take strength and frequency into
> account. As others
> > have mentioned, he wanted to increase hand alternation.
> This is why the
> > vowels are placed on the left hand. He also wanted to
> promote typing
> > from the outside-in, as mentioned, since this is easier than
> typing from
> > the inside towards the outer end.
> >
> > He thus wanted to make the most common digrams easy to
> type. TH is the
> > most common digram in English. It is typed on the home
> row in Dvorak.
> > Likewise, OU is the most common vowel-vowel digram in
> English. This is
> > probably the reason Dvorak put the U on the left index
> finger instead of
> > putting I there. I is about twice as frequent as U in
> English, but the
> > vowel-vowel digrams involving U are about 40% more common
> than the
> > vowel-vowel digrams involving I. Thus Dvorak slightly
> sacrificed the
> > concentration of keys in the home keys/home row in order
> to facilitate
> > typing the most common digrams.
> >
> > The frequency of the following U digrams [au ua ou uo eu
> ue ju uj qu uq]
> > exceeds the frequency of the following I digrams [ai ia oi
> io ei ie ji
> > ij qi iq] by about 40% using a sample of english texts.
> >
> > In other words, I is more common than U, but I is most
> often combined
> > with consonants (which are on the other hand in Dvorak),
> whereas U is
> > most often combined with other vowels. Some people have
> tried to
> > "improve" Dvorak by swapping the I and the U, but this is
> questionable
> > when one takes into account digrams.
> >
> > For comparison, a layout like Colemak <http://colemak.com/
> <http://colemak.com/>
> > has a
> > slightly higher percentage of keys in the home row than
> does Dvorak, and
> > at first sight, appears slightly better than Dvorak. But, it
> places H
> > and U off the home keys, which slows down the typing of
> the TH and OU
> > digrams (as well as HE, which is the second most common
> digram in
> > English). There are also other very common digrams
> involving H, such as
> > CH, SH, and WH. Those involving U include QU (question
> ), JU, AU, etc.
> >
>
>
>