Colemak versus Dvorak

Please compose all posts in Emacs.

Moderators: phlip, Moderators General, Prelates

lalop
Posts: 210
Joined: Mon May 23, 2011 5:29 pm UTC

Re: Colemak versus Dvorak

Postby lalop » Tue May 14, 2013 3:52 am UTC

In terms of ease of learning: I've recently discovered the Tarmak Transitional Layouts which are, imo, the best way to learn qwerty -> colemak. For a task as huge (and, for many, insurmountable) as switching keyboard layouts, it makes a lot of sense to split it up into smaller, more manageable steps.


As for workman, I think I would only recommend it if his ergonomic argument holds up experimentally for you. Although I, too, found it theoretically compelling, I soon realized that the middle column home row keys are not actually difficult for me at all, one of the main "issues" he had been optimizing for. Meanwhile, his layout seems to have sacrificed several other metrics, most notably consecutive finger use (the biggest annoyance in typing). As well as worsening many of the shortcuts.

I ended up making a Colemak vs QWERTY shortcut diagram, which expresses my claims about colemak's shortcuts in a much clearer fashion:

Image

green - unchanged shortcuts
dark purple - improved shortcuts
light purple - improved shortcuts that we probably won't actually use
orange - equally good shortcut (but requires relearning)
red - deproved shortcuts

Morals:
  1. Colemak has an exceptionally high shortcut density on the left side, higher even than QWERTY.
  2. Ctrl should be moved to Caps Lock to take full advantage of this. (IMO, making it Backspace by default was a serious miscalculation on Shai's part.)
Last edited by lalop on Tue May 14, 2013 4:44 pm UTC, edited 1 time in total.

User avatar
cjmcjmcjmcjm
Posts: 1158
Joined: Tue Jan 05, 2010 5:15 am UTC
Location: Anywhere the internet is strong

Re: Colemak versus Dvorak

Postby cjmcjmcjmcjm » Tue May 14, 2013 4:53 am UTC

lalop wrote:In terms of ease of learning: I've discovered the Tarmak Transitional Layouts which are, imo, the sanest way to transition qwerty -> colemak. For such a big (and, for many, insurmountable) task like switching keyboard layouts, it makes sense to split it up into smaller, more manageable steps.


As for workman, I think I would only recommend it if you find his ergonomic argument holds up experimentally for you. Although I, at first, found it theoretically compelling, I soon realized that the middle column home row keys are not actually difficult for me at all, one of the main thrusts of his argument. Meanwhile, his layout seems to have sacrificed several other metrics, most notably consecutive finger use (easily the biggest annoyance in typing). As well as worsening many of the shortcuts.

I ended up making a Colemak vs QWERTY shortcut diagram, which expresses what I've been saying in much clearer fashion:

Image

green - unchanged shortcuts
dark purple - improved shortcuts
light purple - improved shortcuts that we probably won't actually use
orange - equally good shortcut (but requires relearning)
red - deproved shortcuts

The morals:
  1. Colemak has an exceptionally high shortcut density on the left side, higher even than QWERTY.
  2. Ctrl should be moved to Caps Lock to take full advantage of this. (IMO, making it Backspace by default was a serious miscalculation on Shai's part.)

I've got a Mac, so I can use the control on the left. Honestly, I wish I could remap command to both control and alt in Windows.
frezik wrote:Anti-photons move at the speed of dark

DemonDeluxe wrote:Paying to have laws written that allow you to do what you want, is a lot cheaper than paying off the judge every time you want to get away with something shady.

lalop
Posts: 210
Joined: Mon May 23, 2011 5:29 pm UTC

Re: Colemak versus Dvorak

Postby lalop » Wed May 15, 2013 10:29 pm UTC

Not sure how that would square with emacs. But anyway, use tarmak (assuming you already touchtype qwerty), transition through the layouts at your own pace. Post updates in hopefully a month or so, possibly in the colemak forum.

Tarmak promises to make keyboard transitions unprecedentedly easy - in the process, possibly skewing this debate firmly in the direction of colemak - but more feedback is required to see if this actually pans out.

User avatar
Copper Bezel
Posts: 2416
Joined: Wed Oct 12, 2011 6:35 am UTC
Location: Web exclusive!

Re: Colemak versus Dvorak

Postby Copper Bezel » Wed May 22, 2013 7:16 am UTC

Odd question, but is the chart based on web browser use (or maybe a specific IDE or editor) rather than just common shortcuts broadly? If you add office suites to the mix, I, O, and N should be included (purple) while B should be green and L red. (Italics, open, new, bold, left-justify.)
So much depends upon a red wheel barrow (>= XXII) but it is not going to be installed.

she / her / her

lalop
Posts: 210
Joined: Mon May 23, 2011 5:29 pm UTC

Re: Colemak versus Dvorak

Postby lalop » Thu May 23, 2013 3:44 pm UTC

Eh, the chart's values are common shortcuts on web browsers, file browsers, most editors, and of course zxcv most anywhere.

Copper Bezel wrote:If you add office suites to the mix, I, O, and N should be included (purple) while B should be green and L red. (Italics, open, new, bold, left-justify.)


I seriously doubt those shortcuts can be considered common (especially "left justify", what?). However, this is the resulting map:

Image

Here's the keyboard layout analyzer template, so that you can make your own custom shortcut charts.

User avatar
Copper Bezel
Posts: 2416
Joined: Wed Oct 12, 2011 6:35 am UTC
Location: Web exclusive!

Re: Colemak versus Dvorak

Postby Copper Bezel » Thu May 23, 2013 7:22 pm UTC

Yeah, sorry, I guess some of the shortcuts I think of as everyday necessities are really just as arcane as something from Emacs. Ctrl + L seems normal to me simply because centering a heading or title with Ctrl + E, then returning to normal text, seems normal to me.

Thanks for the link. I'm honestly not really thinking very hard about switching from Qwerty (International,) but I do like the look of Colemak.
So much depends upon a red wheel barrow (>= XXII) but it is not going to be installed.

she / her / her

JapanYoshiTheGamer
Posts: 0
Joined: Thu May 02, 2013 12:18 am UTC

Sorry if I derailed the thread.

Postby JapanYoshiTheGamer » Sat Oct 12, 2013 1:02 pm UTC

Anyone annoyed by how iOS doesn't have neither Colemak nor Dvorak? I'm a Dvorak user myself (if anyone read my post from eons ago) actually went as far as to start a petition. I tried to post the URL but the bot flagged it for spam, so can you PM me? Thanks!

lalop
Posts: 210
Joined: Mon May 23, 2011 5:29 pm UTC

Re: Sorry if I derailed the thread.

Postby lalop » Tue Oct 22, 2013 11:32 am UTC

That's why you don't use iOS.

Petitions are rather pointless. Apple will continue to do what it wants, and the approach they've been using is "our way or the highway".

Anyway, typewriter keyboards don't really make much sense on a touchscreen in the first place - we discuss this in some detail. As such, I wouldn't miss colemak or dvorak that much, but rather the more touchscreen-optimized keyboards like messagease, KALQ. Messagease doesn't even really require autocorrect!

Sorry I couldn't give a more encouraging answer, but when you buy into a system that's famous for locking out alternatives..


Edit (June 2014): It looks like they're finally going to allow custom keyboards in iOS 8. I was expecting to lol as yet another year passed without any change, but this outcome is good as well!
Last edited by lalop on Tue Jun 03, 2014 11:47 am UTC, edited 1 time in total.

BoneHead
Posts: 7
Joined: Wed May 21, 2014 10:43 pm UTC

Re: Colemak versus Dvorak

Postby BoneHead » Tue May 27, 2014 11:56 pm UTC

Neo of course. Just look at layer 3 (reached with the Mod3 modifier key, which is colored dark grey):

Image

As for the actual layout, it's optimized for German, but should still be superior to QWERT for English, although in that aspect probably not as great as Dvorak or Colemak. Looks like this:

Image

The six layers it has are:

1 Lowercase characters (no modifiers)
2 Uppercase characters, typographical characters (shift)
3 Special characters for programming, etc. (mod3)
4 WASD-like movement keys and number block (mod4)
5 Greek characters (shift+mod3)
6 Mathematical symbols and Greek uppercase characters (mod3+mod4)

lalop
Posts: 210
Joined: Mon May 23, 2011 5:29 pm UTC

Re: Colemak versus Dvorak

Postby lalop » Sun Jun 01, 2014 7:45 am UTC

I hope people aren't regularly using Neo to type English; it has even more frequent same-finger usage than QWERTY!

Code: Select all

Keyboard Layout  Same-Finger¹ Period
        Colemak        1.72%      58
         Dvorak        2.36%      42
         QWERTY        4.95%      20
  Neo (English)        5.86%      17

the worst offenders being:

Code: Select all

digraphs  Frequency   Period
   al/la      0.88%      114
   ce/ec      0.61%      165
   co/oc      0.50%      200
   ow/wo      0.38%      262
   ep/pe      0.37%      270
   op/po      0.34%      290
   ew/we      0.31%      326
   iv/vi      0.28%      353
   ns/sn      0.28%      360
   hs/sh      0.22%      456
   gr/rg      0.17%      592
   ...


Extra layers are always great, but hardly exclusive to Neo. Furthermore, Neo's layers have some issues that significantly limit their usefulness abroad:

  1. An ISO keyboard is pretty much required for those extra Mod keys. On an ANSI keyboard, you only have one easily reachable extra Mod key (the Caps Lock), making the Neo layers unacceptably wasteful.2
  2. Some of Neo's Mod3 placements are themselves questionable in English (I'm assuming they make more sense in German). Most obviously, the following keys:

    Code: Select all

    Letter  Frequency  Period
         =      0.10%     988
         :      0.10%    1034
         /      0.08%    1318
         ?      0.06%    1770
         >      0.05%    2047
         {      0.04%    2322
         }      0.04%    2330
         [      0.04%    2703
         \      0.03%    3838
         @      0.01%    7924

    are all prioritized over " (0.25%) and ' (0.24%). (They're the only way to type those in Neo, so this isn't exactly a redundant optimization.)

Instead of Neo, I'd stick with the usual approach of choosing the base layout then adding extra layers as needed. (As a bonus, that second step is optional.) In the extreme case, the Neo files might even be used as a template (though, again, only really useful if you have an ISO keyboard).

Finally, and for similar reasons, I think there's a real argument for customizing those layers on your own:

  1. Doing so makes it easier to learn (seeing as you choose the keys).
  2. Peoples' usecases for misc punctuation and movement/control keys are much more liable to vary, making a one-size-fits-all solution unlikely.
  3. The data on such usecases is often limited (or plain unavailable, since movement/control are normally discarded), making it, again, less likely for others to improve over the informal approach.

Full disclosure: for me, both approaches are out, since I use Caps Lock as Ctrl to take advantage of Colemak's pretty darn good shortcuts. Others, like DreymaR, do use Caps for their "Extend layer".




1. Keys absent from the other [base] layouts not included. Because some of Neo's keys are duplicated, a perfect typist might be able to score slightly better, often at the cost of farther reach.
2. For example, Neo spreads both important and unimportant punctuation and navigation over two distinct layers. To avoid having to hit Caps+Shift, however, ANSI extensions need to fit important punctuation/navigation into the single Caps layer.

ajh
Posts: 11
Joined: Thu May 29, 2014 3:26 pm UTC

Re: Colemak versus Dvorak

Postby ajh » Sun Jun 01, 2014 9:56 am UTC

This kind of ruins my latest efforts. Having touch-typed in qwerty for some years, I switched to Neo the hard way maybe a month ago. I’m reconsidering the other candidate; is the German Dvorak any good for English texts?

lalop
Posts: 210
Joined: Mon May 23, 2011 5:29 pm UTC

Re: Colemak versus Dvorak

Postby lalop » Tue Jun 03, 2014 11:30 am UTC

ajh: I'm sorry to hear that. T_T

My original suspicion was that Neo's frequent same-finger usage was due to being optimized for German, rather than English. However, preliminary results1 now suggest otherwise; Neo appears to have relatively high same-finger when typing German as well:

Code: Select all

Preliminary Results (German bigrams): Analyzing English-Letter Bigrams Only

      Keyboard Layout  Same-Finger  Period
               Dvorak        3.38%      30
German Dvorak Type II        3.70%      27
             Colemak²        4.83%      21
              Neo 2.0        7.37%      14
               QWERTY        9.63%      10

[I stress that these results are preliminary and incomplete. For one thing, the non-English letters are not yet counted, and it may be some time before I'm able to include them. Furthermore, punctuation is not included in this dataset, making the numbers unrepresentative of all text. However, I've also tried inputting sample German text into patorjk's analyzer, and it came up with these same rankings.]


ajh: Bearing in mind that there's a chance this preliminary analysis is wrong or misleading, I would still take another look at Dvorak (German or perhaps even English) as a possible base layer. Ideally, you may be able to re-add extra layers as well; the "Caps Lock" key is free in both layouts.

As for your question about German Dvorak's English typing (images for reference):
Spoiler:
Image
Dvorak

Image
German Dvorak Type II (image credit Ulf Stegemann)
  1. Same-finger increased from English Dvorak's 2.36% to 2.80% - a bit much by my standards, but still better than Neo's 5.86%.
  2. The more significant changes, I think, were the swaps themselves. The I/U swap can be ignored for this purpose (some analyzers even suggest it to be an improvement), but the T/R swap put the very common T on the relatively weak ring finger, increasing its frequency from 11.14% (once every 9 keys on average) to 13.59% (once every 7 keys on average). The TH/HT bigram (2.13%) was made more difficult, though DN/ND (0.77%) is now closer.
I've probably missed a few details, but I hope that was helpful. Though I think it's reasonable to suggest that the English typing was worsened, it's difficult to say by how much.




1. I used the German bigram data from this site to try to estimate same-finger. Alternative sources, especially those including punctuation, would be appreciated.
Because some of you probably want to see the culprit bigrams by this point, and so that my numbers may be double-checked, here are the same-finger bigrams in more detail:
Spoiler:

Code: Select all

Preliminary Results (German same-finger bigrams): Analyzing English-Letter Bigrams Only

Neo 2.0:
digraphs  Frequency   Period
   al/la      1.02%       98
   ns/sn      0.76%      131
   ew/we      0.68%      147
   gr/rg      0.45%      223
   ez/ze      0.44%      225
   hn/nh      0.34%      295
   kn/nk      0.31%      322
   ft/tf      0.25%      396
   bn/nb      0.24%      410
   ms/sm      0.24%      420
   ep/pe      0.24%      422
   hs/sh      0.23%      429
   mn/nm      0.23%      433
   co/oc      0.20%      507
   bs/sb      0.20%      512
   ce/ec      0.19%      522
   op/po      0.18%      571
   ow/wo      0.17%      593
   iv/vi      0.17%      600
   hm/mh      0.16%      619
   ks/sk      0.14%      696
   bm/mb      0.09%     1119
   wz/zw      0.08%     1221
   oz/zo      0.07%     1335
   eo/oe      0.07%     1438
   bh/hb      0.05%     1843
   hk/kh      0.04%     2252
   km/mk      0.04%     2269
   bk/kb      0.01%     6692
   dy/yd      0.01%     8776
   dj/jd      0.01%    10491
   pz/zp      0.01%    14758
   ux/xu      0.01%    16207
   cp/pc      0.00%    21020
   dq/qd      0.00%    23365
   pw/wp      0.00%    26806
   cz/zc      0.00%    34201
   cw/wc      0.00%    38325
   jy/yj      0.00%   252055
   qy/yq      0.00%  1740425
   jq/qj      0.00%  7208376
                           
   Total      7.37%       14


QWERTY:
digraphs  Frequency  Period
   de/ed      2.74%      37
   nu/un      1.42%      70
   rt/tr      0.85%     118
   gr/rg      0.45%     223
   lo/ol      0.39%     258
   br/rb      0.38%     260
   mu/um      0.35%     282
   fr/rf      0.34%     291
   hn/nh      0.34%     295
   gt/tg      0.29%     342
   ft/tf      0.25%     396
   mn/nm      0.23%     433
   ce/ec      0.19%     522
   hm/mh      0.16%     619
   ik/ki      0.16%     627
   hu/uh      0.15%     678
   bt/tb      0.14%     738
   rv/vr      0.13%     779
   sw/ws      0.12%     825
   tv/vt      0.08%    1238
   az/za      0.07%    1338
   fg/gf      0.06%    1732
   jn/nj      0.05%    1880
   bg/gb      0.05%    1893
   ju/uj      0.04%    2276
   gv/vg      0.04%    2607
   jm/mj      0.02%    4025
   bf/fb      0.02%    4219
   cd/dc      0.02%    4297
   fv/vf      0.01%    7387
   my/ym      0.01%    8026
   ny/yn      0.01%    8695
   hj/jh      0.01%   11827
   bv/vb      0.01%   12826
   hy/yh      0.01%   13064
   sx/xs      0.00%   23086
   uy/yu      0.00%   29063
   aq/qa      0.00%   44823
   wx/xw      0.00%   82393
   jy/yj      0.00%  252055
   qz/zq      0.00%  599911
                           
   Total      9.63%      10

Colemak:
digraphs  Frequency  Period
   cs/sc      0.78%     128
   eu/ue      0.52%     192
   dt/td      0.34%     295
   hn/nh      0.34%     295
   kn/nk      0.31%     322
   gt/tg      0.29%     342
   hl/lh      0.23%     430
   mn/nm      0.23%     433
   rw/wr      0.18%     543
   hm/mh      0.16%     619
   ln/nl      0.16%     636
   kl/lk      0.14%     704
   bt/tb      0.14%     738
   fs/sf      0.13%     796
   dg/gd      0.12%     828
   tv/vt      0.08%    1238
   lm/ml      0.07%    1337
   az/za      0.07%    1338
   pt/tp      0.07%    1341
   bd/db      0.06%    1699
   jn/nj      0.05%    1880
   bg/gb      0.05%    1893
   dp/pd      0.05%    2212
   hk/kh      0.04%    2252
   km/mk      0.04%    2269
   gv/vg      0.04%    2607
   dv/vd      0.04%    2844
   jm/mj      0.02%    4025
   cf/fc      0.01%    9746
   gp/pg      0.01%   10749
   hj/jh      0.01%   11827
   bv/vb      0.01%   12826
   bp/pb      0.00%   21608
   jl/lj      0.00%   22702
   iy/yi      0.00%   24148
   pv/vp      0.00%   24305
   aq/qa      0.00%   44823
   rx/xr      0.00%   62158
   jk/kj      0.00%   71891
   wx/xw      0.00%   82393
   qz/zq      0.00%  599911
                           
   Total      4.83%      21

Dvorak:
digraphs  Frequency  Period
   nr/rn      0.41%     246
   ls/sl      0.31%     326
   tw/wt      0.20%     495
   dh/hd      0.17%     586
   hm/mh      0.16%     619
   ik/ki      0.16%     627
   nv/vn      0.15%     647
   ip/pi      0.14%     714
   ku/uk      0.14%     734
   df/fd      0.13%     751
   rv/vr      0.13%     779
   dg/gd      0.12%     828
   dm/md      0.12%     853
   ej/je      0.10%    1051
   sz/zs      0.09%    1097
   pu/up      0.09%    1119
   bm/mb      0.09%    1119
   gm/mg      0.08%    1308
   fm/mf      0.07%    1436
   gh/hg      0.07%    1532
   bd/db      0.06%    1699
   fg/gf      0.06%    1732
   bh/hb      0.05%    1843
   bg/gb      0.05%    1893
   lz/zl      0.05%    2085
   iu/ui      0.04%    2316
   fh/hf      0.04%    2360
   bf/fb      0.02%    4219
   ct/tc      0.02%    5157
   ix/xi      0.01%    6859
   px/xp      0.01%    8027
   kp/pk      0.01%   10535
   py/yp      0.01%   12854
   ux/xu      0.01%   16207
   ky/yk      0.00%   20573
   iy/yi      0.00%   24148
   uy/yu      0.00%   29063
   cw/wc      0.00%   38325
   kx/xk      0.00%   47191
   xy/yx      0.00%  197043
   oq/qo      0.00%  219010
                           
   Total      3.38%      30

German Dvorak Type II:
digraphs  Frequency  Period
   nt/tn      0.71%     141
   ls/sl      0.31%     326
   rw/wr      0.18%     543
   dh/hd      0.17%     586
   hm/mh      0.16%     619
   ik/ki      0.16%     627
   nv/vn      0.15%     647
   ip/pi      0.14%     714
   ku/uk      0.14%     734
   df/fd      0.13%     751
   dg/gd      0.12%     828
   dm/md      0.12%     853
   cr/rc      0.10%     971
   ej/je      0.10%    1051
   sz/zs      0.09%    1097
   pu/up      0.09%    1119
   bm/mb      0.09%    1119
   tv/vt      0.08%    1238
   gm/mg      0.08%    1308
   fm/mf      0.07%    1436
   gh/hg      0.07%    1532
   bd/db      0.06%    1699
   fg/gf      0.06%    1732
   bh/hb      0.05%    1843
   bg/gb      0.05%    1893
   lz/zl      0.05%    2085
   iu/ui      0.04%    2316
   fh/hf      0.04%    2360
   bf/fb      0.02%    4219
   ix/xi      0.01%    6859
   px/xp      0.01%    8027
   kp/pk      0.01%   10535
   py/yp      0.01%   12854
   ux/xu      0.01%   16207
   ky/yk      0.00%   20573
   iy/yi      0.00%   24148
   uy/yu      0.00%   29063
   cw/wc      0.00%   38325
   kx/xk      0.00%   47191
   xy/yx      0.00%  197043
   oq/qo      0.00%  219010
                           
   Total      3.70%      27

Probably the strangest one (to me) is al/la, a common bigram both in German and English. I'm unable to phantom why those two letters were put on the same finger.

2. Funnily enough, if you type C with your index finger, Colemak's result actually lowers to 4.10%, due to cs/sc (0.78%) being a common German bigram. This is the opposite to English, where doing so increases your same-finger by 0.18%.
Last edited by lalop on Tue Jun 03, 2014 9:10 pm UTC, edited 2 times in total.

User avatar
PeteP
What the peck?
Posts: 1451
Joined: Tue Aug 23, 2011 4:51 pm UTC

Re: Colemak versus Dvorak

Postby PeteP » Tue Jun 03, 2014 11:54 am UTC

Question in your analysis of special characters, did you include code in your analysis? Because I would say I use the ones you listed more than " and ' (well maybe not when I am working with many strings) and they are a reason I like neo because I find their position convenient. (Though I don't care about typing two with same finger anyway, I just like having special characters in easy reach.)

lalop
Posts: 210
Joined: Mon May 23, 2011 5:29 pm UTC

Re: Colemak versus Dvorak

Postby lalop » Tue Jun 03, 2014 5:04 pm UTC

I'm using MTGAP's dataset for the English analysis. Narrowed-down datasets (including Java, C) can be found in the freq_types folder.

Even for those two languages, the data still indicates more instances of " than all the other listed special characters except =. And that's not even getting into the other 54% of time doing non-coding work.1

PeteP wrote: special characters...are a reason I like neo because I find their position convenient. (Though I don't care about typing two with same finger anyway, I just like having special characters in easy reach.)

To quote that overused commercial, why don't we have both? The main source of Neo's same-finger is the base layout, after all, which can be replaced without affecting any of the special layers.


1. Though, if this comic is accurate, it might be more like 37%.

User avatar
PeteP
What the peck?
Posts: 1451
Joined: Tue Aug 23, 2011 4:51 pm UTC

Re: Colemak versus Dvorak

Postby PeteP » Wed Jun 04, 2014 5:53 pm UTC

I might actually do that and revert to qwerty, because others using my computer is inconvenient with a non standard layout. (It is easy to change the layout back but I also put keyboard stickers on them.) I just need to make my own stickers which show qwerty + layers.


Return to “Religious Wars”

Who is online

Users browsing this forum: No registered users and 2 guests