Looking for feedback on an ergonomic layout (PYUW)

[william]

Hey guys,
So I have been playing around with ideas for a keyboard layout for maybe two years now, and this is the newest iteration. I'd like to get some feedback on what you think about it.

It is heavily influenced by Dvorak, with the vowels and punctuation on one side.

My priority for fingers goes middle>index>ring>pinky, and my priority for rows is home>top>bottom. The most difficult keys for me to reach on this layout are the top row keys toward the very middle, so I gave those the lowest priority. I didn't make any concessions to QWERTY for two reasons; I think it makes it difficult to switch between two layouts when they are too similar, and because QWERTY is garbage. As far as shortcut keys are concerned, those can go where you are used to them by adding them to an alternate layer.

I ran some books through a text analyzer for key frequency data, so here is that:

Code: [Select]

Key Occurrences %
E 120465 11.84
T 95755 9.41
A 84088 8.26
O 81596 8.02
N 71332 7.01
I 68475 6.73
H 65791 6.47
S 61514 6.05
R 51862 5.10
D 49418 4.86
L 41038 4.03
U 31742 3.12
W 28699 2.82
M 25420 2.50
Y 22914 2.25
C 21904 2.15
G 21900 2.15
F 19495 1.92
B 15922 1.56
P 14727 1.45
K 10772 1.06
V 8566 0.84
J 1820 0.18
X 1178 0.12
Q 857 0.08
Z 424 0.04

Here is a link to a layout analyzer I used as a reference:
http://patorjk.com/keyboard-layout-analyzer/#/load/4ccDHpcF
The top scoring layout (JYUP) is one I made a while back and is what I've been using for a while. I'm pretty happy with it, though I feel 'W' could be more convenient, and going from 'Q' to 'U' is a bit annoying, but it's not like that is a big consideration, as you can see in the text analysis data above. I don't fully understand the weighting of the metrics he used, but the fact that this is for staggered keyboards probably has a lot to do with JYUP scoring better than PYUW.

I use two different keycap profiles to help me keep my hands oriented. It was a lot more difficult to get used to the layout before I added them.
Here is a picture of my keyboard:


I think eventually I would like to ditch all the f keys and stuff in the middle in favor of having a large track pad fill that space instead.

[Oobly]

The layout looks very good, particularly for single letter frequencies. The only issues that I can see immediately are "ND" and possibly "WA". What language is your letter frequency chart / layout for? If it's for English, the letter "C" should be higher in the list, somewhere near the "DLU", but of course this depends on the text (corpus) used for analysis.

Here is a site with interesting information on English letter, bigram and n-gram frequencies: http://norvig.com/mayzner.html

A very good layout analyser is the AdNW software. It's flexible in that you can assign your own weighting to various factors and let it optimise your layout for a particular corpus: http://www.adnw.de/index.php?n=Main.Downloads

In the "Entwicklungsprogramme" section there is the actual analyser and a couple other tools.

You can create a matrix-like layout in the config section of the patorjk analyser, too, by selecting and loading one of the ErgoDox layouts. It's a very handy tool for doing a quick analysis of small changes in a layout to see if they are actually improvements or not.

Here are some results with my layout also in the mix http://patorjk.com/keyboard-layout-analyzer/#/load/xxWlDBk7

I suspect the downfall of PYUW in the analysis is consecutive finger use on the right index finger. Seems to be the weakest metric in the results for that layout.

[william]

Hey, thanks for the links; they sound like they might be really helpful.

As for where the text analysis was from, it was Alice's Adventures in Wonderland, Adventures of Huckleberry Finn, and The Adventures of Sherlock Holmes, in their entirety, so I wouldn't have thought that the frequency of 'C' could be all that far off. It doesn't surprise me that 'I' is lower in my analysis, since most books would probably be from more of a first person perspective than those. I'll have to look into it more.

I think you are right about ND. R and D are swapped in my original JYUP layout, which I put more consideration into bigrams with. Guess I should swap them back. Also wasn't expecting S to pull that far ahead of H. Not sure If I should move it though, cause i like TH being a finger roll. Then again, not sure if I should be looking to change anything in the first place, since I am mostly happy with my old layout and it has taken a good bit of time to start to feel at all comfortable on it.

I didn't realise they had other keyboard layouts in patorjk, that is pretty awesome. How did Q end up in such a strong spot in your layout? Is english your primary language?

thanks again for the info.

[batfink]

Agree, it's not bad from the point of view of common letters in easy-to-reach places, and the left/right hand balance is good.

But it falls down on same finger bigrams. It's barely better than Qwerty in that regard. In order of worst offenders: ND, WA, YO, LD. 

[william]

At what point do you thing consideration for bigrams should fall off of the spectrum? According to the data linked by Oobly, WA happens about once in every 300 words. I imagine YO is a way more frequent bigram in every day communication than it is in books, but even so, how do you determine the weight that a consideration like that should be given versus choosing the position solely on individual key frequency?

I hope that question doesn't seem smarmy, because I am curious to hear what others think about bigrams in particular.

As far as being barely better than QWERTY for bigrams goes, that is the one area where QWERTY should be expected to do well, since the whole point of the layout in the first place was to prevent typewriter arm jamming, so I don't feel that bad about my layout being only slightly better than it, in that one regard.

[batfink]

As far as being barely better than QWERTY for bigrams goes, that is the one area where QWERTY should be expected to do well, since the whole point of the layout in the first place was to prevent typewriter arm jamming, so I don't feel that bad about my layout being only slightly better than it, in that one regard.

The design of early typewriters to prevent jamming is a not quite the same issue of same-finger bigrams. Especially when it comes to index fingers, where there are six keys to consider. If it was ever Qwerty's claim that its rate of same finger bigrams is low on account of the need to avoid jamming, then its design fails quite badly even on its own terms!

I make it that using this layout, around 6% of bigrams typed will be same-finger. Qwerty is also around the 6-7% mark. That doesn't compare well with, for example, Colemak which is less than 2%.

Of course, the amount of bigrams on the same-finger you are willing to live with is a matter of personal taste. But personally I would say it's the second most important factor (having common keys in easy-to-reach positions being the first).