• You are not logged in.

    Optimizing Colemak for Finnish

    • Started by vvanhala
    • 6 Replies:
    • Reputation: 0
    • Registered: 16-Oct-2018
    • Posts: 3

    Hello everyone!

    Here's a new user trying to enter the world of alternative layouts so I'll introduce myself a bit first.
    (If you are in a hurry here's the tl:dr: How do I relocate the letter J without ruining typing in English?)

    I've been touch typing qwerty since my teens so for two decades now. I recently started deliberately practicing to improve my speed and proficiency with numbers and special characters. During the process I inevitably came across the alternative layout community and started to learn more. What a world of intriguing optimizing! I must say I'm a bit baffled I never did this before since I love optimizing everything and learning new skills. I guess I thought the barrier to learn a new system would be too great. Reading the experiences here assured me it would most certainly be worth it.

    My research with different layouts led me to Colemak and I've been reading this forum to understand all the different flavours. I'm now set with Colemak-CAW as my preferred new layout to learn.

    The only difficulty is modifying it to suit Finnish. I read DreymaR's suggestions on localization and I'll most probably fit the Finnish umlauts ä and ö for the right pinky as they are in the finnish qwerty. I'll have to experiment with that a bit but I'm sure it's going to go fine. I didn't find anyone who would have written extensively about their experience with Colemak for Finnish. There are layouts designed purely for Finnish, but it became clear Finnish and English are so different in their common letters that it won't be possible to find a layout thats optimal for both. So I'm trying to find how to tweak Colemak to be a bit more in favor of Finnish.

    I did a bit of writing and the only major problem is with the letter J. It has a horrible location on the Colemak. It's located in the most strenuous finger location on the keyboard and in Finnish it's quite common.

    Here are the most common letters for English and Finnish. [according to practicalcryptography.com]

    Screen-Shot-2018-10-16-at-13-58-50.png

    The letters W and F have very good spots on the keyboard and they practically don't exist in Finnish. I'm temped to fit J on either of them but I can't just swap J for W for example because that would create a horrible WH bigram. Exchanging J for F would create a difficult FL bigram (floor, flour, flee....). So a simple swap just wont do it. One possbile solution would be to just swap B and J. That would improve J slightly in Finnish but without sacrificing much in English.

    So here's the question: How do I relocate the J into a better place without messing up writing in English? I took a look at the most common letters and bigrams on both languages but it's too daunting to try to do without some sort of algorithm. I simply don't know enough about keyboard layout design to make an educated decision.

    -Valtteri

    Last edited by vvanhala (16-Oct-2018 19:33:56)
    Offline
    • 0
    • Reputation: 117
    • From: UK
    • Registered: 14-Apr-2014
    • Posts: 978

    Well, although J is more common than F and W in Finnish, according to the practical cryptography website you quote, the frequencies (percent) are:
    J :  2.04
    F :  0.19
    W :  0.09

    I know J is particularly annoying on a standard keyboard due to the stagger, but still - at 2%, it's still not that common, so I'd have to wonder whether it's worth the effort of moving it?

    Last edited by stevep99 (16-Oct-2018 16:30:55)

    Using Colemak-DH with Seniply.

    Offline
    • 0
    • Reputation: 0
    • Registered: 16-Oct-2018
    • Posts: 3

    That is a fair point Steve. I understand not messing with the layout and just learning to live with it. On the other hand if I'm going to invest in learning a completely new way of writing I'd like for it to not have any major annoyances.

    As for the statistics it looks quite a bit different if we look at the most common words in Finnish (again form practicalcryptography.com)

    1. JA :  3.94 (and) <- look at the percentage here
    2. ON :  2.86 (is)
    3. OLI :  1.05 (was)
    4. HÄN :  0.81 (he/she)
    .
    7. JOKA :  0.53 (which)
    .
    17. JÄLKEEN :  0.24 (after)
    .
    20. JONKA :  0.20 (whose)
    .
    24. JOSSA :  0.18 (in which)

    So even if the single letter J percentage looks small it is very prominently in the most common word "ja" (and) and many other common words. "Ja" is also the 20th most common bigram at 1% of all bigrams.

    I tried swapping J and B and that feels much better already.

    Last edited by vvanhala (16-Oct-2018 19:57:34)
    Offline
    • 0
    • Reputation: 23
    • From: Belgium
    • Registered: 26-Feb-2008
    • Posts: 482

    As a Flemish (Dutch) speaker, we also have relatively frequent "J" (eg. ja = yes, je/jij/jou/jouw = you/your, ...), so I also initially considered moving it to a better spot.  But I quickly realized that:

    a) There is no evident spot to move it to without breaking/impairing other aspects of the layout

    b) For many non-English speakers, especially in a technical or business environment, their native language constitutes only a part of their typing.  For me, I estimate I type at least 50% English (for example right now) and command line fu, so improving some letters for your native language may have less benefit than you'd think at first sight.  (Very likely, moving eg. Tab, Ctrl or dash would benefit me a lot more than moving "J"...)

    c) Moving away from the standard makes using Colemak harder, esp. for Linux and Mac users who can otherwise benefit from built-in implementations.

    Offline
    • 0
    • Reputation: 214
    • From: Viken, Norway
    • Registered: 13-Dec-2006
    • Posts: 5,364

    I really believe in the Colemak-DH approach. That is, it's quite safe to move keys without changing their fingering like the Cmk-DH mod does. The main reason for this is that bigrams are ever so tricky and Colemak is so close to ideal in this respect that you're almost bound to regret any change of fingering when you're typing English. And most of us do type quite a lot of English after all.

    In that context, we've discussed what people with more common L and J in their languages (such as the Dutch) can do. But as Ghen says, you'll then have to put another letter to fry in the black cauldron!

    Try sorting the letters JLKNMH by frequency and need of access. N and H should win in most relevant languages, but already here we see another interesting fact of Finnish: L is substantially(?) more common than H! Well, well, well. So it seems you don't quite want the standard DH mod then, heh. Also, K is even more common than J whereas they're both rare in English. N and M are about on par between the languages. In English the frequencies are NHLMKJ, in Finnish NLKMJH.

    What to make of that? The extreme approach would be to move H to the Cmk J position since it's less common in Finnish. But that'd ruin your English typing something fierce, regardless of no fingering change! The stretches would be brutal for many very common n-grams. Just don't do that. The DH mod is so good for English typing precisely because it improves those common bigrams so much.

    I'd say that L may warrant a trip to the bottom row though, if you're so inclined. Where Colemak-DH has JL/KN/MH and vanilla Colemak JL/HN/KM, you might use JK/HN/ML(LM?) for instance. That should be an interesting proposition worth trying out.

    For the J problem, I'm afraid I don't have a good solution for you. My suggestion would be to get a non-row-staggered board (curvilinear or other column matrix design) and enjoy its equalizing benefits. Ask in the #hardware channel of the Colemak Discord to find your dream hardware.

    Well, maybe one tiny suggestion. For some, the two-key horizontal stretch is actually easier than the upwards/in one to J. If that's the case for you, you might consider putting Ää on the J position and J on the RightBracket position which on Cmk-CAW sits in the middle of the board! Not sure it'll feel better for you, but there it is. Let us know. Ä is actually more common than J in Finnish so it'll depend on the bigrams for the most part. Also, this trick will break compliance with standard Colemak-CAW which may be annoying in the long run.

    Last edited by DreymaR (17-Oct-2018 08:50:24)

    *** Learn Colemak in 2–5 steps with Tarmak! ***
    *** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***

    Offline
    • 0
    • Reputation: 0
    • Registered: 16-Oct-2018
    • Posts: 3

    Thank you for the insights! This was exactly what I was looking for: experience based recommendations. It's so easy to come up with theories of what could work but only experience tells if it works or not.

    I had to do plenty more testing to see what would work and unsurprisingly you were all correct. It's better to leave the layout as it is and learn to adjust when writing finnish. For example swapping B and J would have created a rather common and uncomfortable bigram BL for the left index finger.

    DreymaR said:

    For the J problem, I'm afraid I don't have a good solution for you. My suggestion would be to get a non-row-staggered board (curvilinear or other column matrix design) and enjoy its equalizing benefits. Ask in the #hardware channel of the Colemak Discord to find your dream hardware.

    Well this just plunged me ever deeper into the rabbit hole. :)

    I did some digging into what is available on the hardware side. The Planck, Lets split and Ergodox all seem like nice options. A Planck would be nice for portability but I'd love to have a thumb cluster and a possibility for split layout. The Dactylus seems pretty ideal but not available for purchase. Maybe I need to start looking for a 3D printer...

    Offline
    • 0
    • Reputation: 214
    • From: Viken, Norway
    • Registered: 13-Dec-2006
    • Posts: 5,364

    Sorry I ruined your life, mate. ^_^

    The ErgoTravel boards look completely awesome to me:
    https://imgur.com/a/JIlV1

    Unlike many of the more popular mini boards, this one has 6×3 keys on the main block, allowing room for locale letters and suchlike. The 5×3 blocks seem too restricted to me.

    Last edited by DreymaR (24-Oct-2018 11:17:05)

    *** Learn Colemak in 2–5 steps with Tarmak! ***
    *** Check out my Big Bag of Keyboard Tricks for Win/Linux/TMK... ***

    Offline
    • 0