Issues regarding keyboard layout

[iandoug]

Hi

As mentioned previously, we have found that what works best is to have vowels on the left and consonants on the right. This is basically the same idea as what Dvorak had, except with a saner allocation of letters -> fingers. The idea is to get a reasonable balance between left hand and right hand usage. Also languages tend to alternate vowels and consonants, with leads to better keyboard action (alternating hands is faster than finger rolls on same hand).

However Dan presents some challenges, in that it is vowel-heavy, made worse by the use of tone characters, which also work better on the left hand (in testing so far).

Compare this analysis of Putin's speech in English, vs Matthew 5-7 in Dan:

Matthew: vowels : 5326 consonants : 6012 (including n and g) tone : 3398 others : 4733 (includes space)

Putin: vowels : 5838 consonants : 9058 tone : 57 others : 3319

As you can see, consonant usage is much higher in English relative to vowels. If we add Dan vowels plus tone together, then its much higher than consonants. This makes getting a balanced keyboard layout challenging.

There is also the problem (well, one of them) inherent in the ISO form factor, especially since the Enter key is so far from right pinky home position. This is over and above the design flaw giving the right pinky the most keys of all fingers. The net result is that no matter what you do, the right pinky is overworked.

There are "cheats" that we can do with ISO, like moving the Enter key to where the AltGr key is, but in truth that leads to awkward hand contortions. The Japanese keyboard would be better in this regard. The only really viable solution is to go with an custom ergo/matrix style layout. This reduces hand travel wherever possible.

Comparisons below of a layout on ISO, and same layout translated as best possible to matrix style. I'll add these layouts to the http://kladan.keyboard-design.com shortly. Also updated the Matthew input text and removed the trailing space at end of each line.

The other issue is the excessive use of the dash tone mark. I've found it works better if you put this on a thumb key, which once again bumps into the unsuitable ISO form factor. Works better on matrix layout.

[HughP]

My thought with some of these tone languages is to have tone markers input with space bar. Like a double tap or triple tap of space bar. Is there a way that we could model that?

[iandoug]

Not with KLA in it's current state. I am aware of some layouts that have that sort of double-tap function (or where tap means one thing and hold means another) but KLA basically just measures key presses and distance fingers move. And it only knows about single presses.

The Matrix layout above is effectively what you want, if you think of splitting the space bar in two, and giving one half to the left thumb and the other half to the right thumb. (once again, exposing fundamental flaws of ANSI/ISO design). That way you just need single tap instead of double tap. Double taps on something as common as /dash/ are going to increase the score dramatically.

[iandoug]

Don't know if you have done this analysis yet (don't remember seeing it), but I did a word count, here's the top thirty (sans tone marks etc). The thing to notice is that they have a LOT of short words (which means the space bar also gets used a lot) and over 1/3 of the top 30 words use letters with diacritics. So am reluctant to force dead keys into the design.

1 : bha : 4904 2 : ö : 4301 3 : nu : 3318 4 : wo : 2566 5 : yö : 2430 6 : ka : 2421 7 : a : 2413 8 : kë : 2335 9 : dhɛ : 2211 10 : gü : 1867 11 : ya : 1644 12 : mɛ : 1481 13 : kö : 1211 14 : sü : 979 15 : pö : 915 16 : wɔn : 860 17 : në : 795 18 : mü : 779 19 : do : 763 20 : dho : 757 21 : A : 625 22 : kwa : 622 23 : yaa : 608 24 : gɔ : 567 25 : dhö : 551 26 : gban : 499 27 : ta : 492 28 : ü : 472 29 : kɔ : 465 30 : dhϋ : 464

[iandoug]

Is there no chance of them dropping bh and dh and going back to ƃ and ƌ ? Suppose it's like English th where we could use ð and þ ...

Only found one occurrence of a Dan word with h preceded by a different letter (l). (and a misspelling of Jonh).

[HughP]

Yea, I was looking at those mis-placed letter h as well. When I remove the French then those are the only two instances of h that do not occur with b/d.

On Tue, Jun 19, 2018 at 2:04 AM, Ian Douglas notifications@github.com wrote:

Is there no chance of them dropping bh and dh and going back to ƃ and ƌ ? Suppose it's like English th where we could use ð and þ ...

Only found one occurrence of a Dan word with h preceded by a different letter (l). (and a misspelling of Jonh).

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/HughP/dnj-corpus/issues/21#issuecomment-398328676, or mute the thread https://github.com/notifications/unsubscribe-auth/AACdJll40uzOQoRdZi6JQHM95NU7nIJyks5t-L6KgaJpZM4UivB_ .

[HughP]

@iandoug I can get KLA to let me re-arrange the keyboard, and then I can copy out the json to a text file, but I can't load any of the files from the list of keyboard layouts.

[iandoug]

Um, working from here (just tested). You on MacOS? Which browser?

[HughP]

I got KLA working... I think I was expecting to be able to export the Personalized keyboards. Thanks for the help. I'll have some analysis here in about two days which might be of some interest. Look for it on a new thread.