Anyone familiar with keyboard optimisation software? I have a request...

[asdfasdf1234]

I have been looking at Carpalx http://mkweb.bcgsc.ca/carpalx/?home fully optimised layouts http://mkweb.bcgsc.ca/carpalx/?full_optimization , and like the ideas, but I have one request —

Their fully optimised layout switches p and ;
I suggest (slightly) more aggressive switching of punctuation.
Acknowledging their model of base effort http://mkweb.bcgsc.ca/carpalx/?typing_effort “BASE EFFORT IS FINGER TRAVEL DISTANCE”



— where qwerty b has an effort of 3.5, qwerty y an effort of 3.0, while qwerty ;: and qwerty /? have an effort of 0.0 and 2.0 respectively, I am interested in layouts that place qwerty ;: on qwerty b and qwerty /? on qwerty y ... and then I have requests that will generate four different layouts.

Option 1:  all letter keys free to migrate (after ;: and /? swap with b and y)
a) ,< and .> in qwerty position
b) ,< and .> shifted one key to the right
Option 2:  zxcv fixed in qwerty position
a) ,< and .> in qwerty position
b) ,< and .> shifted one key to the right

In my opinion those layouts would make moderately good use of standard qwerty keyboards.  Comparing scores for these four layouts would help determine whether it is worth disrupting zxcv, or shifting ,< and .> to the right.

However ... I have been using linux for 18 months, and have no programming experience, and I am not going to dive in to programming and using the Carpalx perl script optimiser if I don't have to.  I wondered if someone here would be able to calculate those layouts easily with their existing experience, using Carpalx default settings, and post them.

If nobody does within a few months then I may have to learn how to run perl scripts and do it myself... haha

Regards

[davkol]

I prefer not to rely on carpalx for layout optimization. The model itself is quite reasonable, but weighs, esp. for triads, are arbitrary and there are some bugs in the reference implementation.

BTW have you looked into MTGAP? I'm not saying it's a big step from carpalx, but some people prefer it at supposedly more user friendly and (arbitrarily) better.

[asdfasdf1234]

I have seen MTGAP, I found his work/site impressive.  As I am asking someone else to calculate this keyboard layout for me I am in no position to judge their software or methodology 

It is the swapping of two uncommon (well, / is used in web addresses and linux command-lines, but that is still fairly uncommon) punctuation keys with two awkward (qwerty) letter keys that I find interesting.

[davkol]

Well, you should care about the corpus and criteria used, esp. if you seek a fully custom layout as opposed something more widespread like DSK or Colemak.

If I understand the OP properly, what you want is basically something of a variation on DreymaR's wide mod (see section 2.1).

[asdfasdf1234]

I experimented with a right-shifted right-hand layout once before, but found it uninteresting because of the excessive reaching of the right hand forefinger.

I am interested in zxcv fixed and zxcv unfixed layouts based on the attached letter-masks (thanks http://www.keyboard-layout-editor.com/#/ for their keyboard layout creator used to create the images).

,< .> in qwerty position


,< .> shifted to the right


Even the single swap of qwerty ;: with b may be interesting (as linux users such as myself may use / a lot)

[davkol]

a right-shifted right-hand layout (…) the excessive reaching of the right hand forefinger.

I don't think so.

Wide mod

• increases the gap between wrists (potentially reducing ulnar deviation),
• makes bottom-row modifiers more accessible (which opens up the option of moving special symbols to a better-accessible layer/Compose),
• makes right Shift, Enter and Backspace more accessible,
• brings the right hand closer to a right-handed mouse (or nav cluster or tenkey respectively),
• reduces number of keys (that are uncommonly used, as you've noted yourself) controlled by the right pinkie… by moving the respective symbols to an area controlled by the longer, stronger and still flexible index fingers (indeed: both of them).

The advantages are overwhelming, if you ask me.

Now, back to the OP:

• Finger travel shouldn't matter much; not nearly as much as motions on longer n-grams. Furthermore, it's calculated pretty naively in carpalx: it assumes that the finger always returns to it's home position, even in same-finger sequences.
• The carpalx model doesn't penalize lateral motions (which is what Colemak ModDH or Workman do). Masking that key can be considered a hack, that kind of forces it to do so in a way, though. That said, esp. the problem with B position on QWERTY can be mitigated by simply (a) using a symmetrical keyboard¹ or (b) using the angle mod on standard keyboards (again, see DreymaR's tips).
• Corner bottom-row keys, such as the QWERTY punctuation spots or ZX, are almost universally considered to be some of the worst on the keyboard. That's why, for example, Colemak doesn't lose much by leaving these parts the same as on US QWERTY.

Yeah, an "academic" debate at best.

¹ EHMAGERD ERGODOX

[jacobolus]

The numbers and heuristics used in most of these layout optimizers are totally arbitrary and not very plausible IMO.

[Pro XKB]

As I am asking someone else to calculate this keyboard layout for me I am in no position to judge their software or methodology

Well, in this case, here are the results from the AdNW optimizer, using the bigram optimization and the corpus that comes with the software:

                 236.641 total effort   186.433 positional effort    left right
                   1.136 same finger rp   7.660 shift same finger top 10.6 11.7
  yplmk /fouq     62.978 hand alternat.  45.989 shift hand alter. mid 29.6 36.2
  hnrtd csaei'     0.912 inward/outward  33.320 inward or outward bot  4.7  7.2
  vzxw; gb,.j     13.784 adjacent        18.807 shift adjacent    sum 44.9 55.1
                  9.1  8.7  9.9 17.2 --.- --.- 14.8 15.9 15.3  9.1 Sh  1.8  1.3


                 236.891 total effort   186.440 positional effort    left right
                   1.114 same finger rp   8.058 shift same finger top 10.6 11.7
  yplmk /fouj     62.978 hand alternat.  45.989 shift hand alter. mid 29.6 36.2
  hnrtd csaei'     0.877 inward/outward  33.342 inward or outward bot  4.7  7.1
  vzxw; gbq,.     13.411 adjacent        18.502 shift adjacent    sum 44.9 55.1
                  9.1  8.7  9.9 17.2 --.- --.- 14.8 14.8 15.3 10.2 Sh  1.8  1.3



                 241.087 total effort   193.757 positional effort    left right
                   1.195 same finger rp   7.134 shift same finger top 11.3 11.7
  wlmpb /fouq     67.113 hand alternat.  29.562 shift hand alter. mid 29.4 35.3
  rnstg dhaei'     0.984 inward/outward  29.125 inward or outward bot  5.7  6.5
  zxcv; yk,.j     12.655 adjacent        21.685 shift adjacent    sum 46.5 53.5
                  9.1 10.8 11.8 14.8 --.- --.- 12.9 15.9 15.3  9.4 Sh  1.4  1.7

                 241.967 total effort   193.765 positional effort    left right
                   1.173 same finger rp   8.313 shift same finger top 11.3 11.7
  wlmpb /fouj     67.113 hand alternat.  29.562 shift hand alter. mid 29.4 35.3
  rnstg dhaei'     0.941 inward/outward  29.147 inward or outward bot  5.7  6.5
  zxcv; ykq,.     12.316 adjacent        20.642 shift adjacent    sum 46.5 53.5
                  9.1 10.8 11.8 14.8 --.- --.- 12.9 14.8 15.3 10.5 Sh  1.4  1.7

I have used the default scoring scheme.  I attach the configuration file, in case you want to modify it.  Which you probably will, as the scheme is completely different from CarpalX.