Home > Remapping > Remapping for Foreign Letters & Symbols > Analysis and Methodology

From: Matthias Schult Date: Fri, Apr 28 2006 7:36 pm

… I have counted the occurrence of characters and character combinations in the Wikipedia in 65 languages using the Latin alphabet and I have counted special characters in some programming languages (C/C++, Java, perl, shell scripts, Python, HTML, css). The goal of course is to create key layouts for the AG to use in other languages. Since I still haven’t got my AG and it’s unlikely that it will arrive tomorrow, I won’t have it this prolonged weekend (1st of May is a holiday). Still I would like to hear your views on my demands for creating such a layout:

A key layout should...

Be as similar as possible for all languages

It is utterly annoying having to type on a slightly different keyboard layout when typing in a foreign language. E.g. I’m used to typing on a German QWERTZ-layout. But when typing Swedish, the å and Å are very hard to reach, because they don’t exist in German, so I have to switch to the Swedish QWERTY-layout frequently mistyping y and z (~0.7% of the Swedish language). Also sometimes one has to cope with a standard US-layout but German keyboard and has to go hunting for things like asterisk, brackets, dash, and so on. So it would be most convenient to have everything but the language’s special characters in the same place. The problem is how different languages should be weighted on creating such a layout. E.g.: All equally? By Wikipedia size? By number of speakers (mother language or total?)? Language families weighted equally? Number of web pages on the internet? The differences are not too big though.

Have the most common characters in the most convenient places

This one is quite obvious, but gets into conflict with (1). The most common characters are language dependent, but not as much as one might think. It only gets problematic with common special characters like ə in Azerbaijani, ë in Albanian, ä in Finnish and Tartar, I in Tartar and Turkish, á in Hungarian, ð in Icelandic, ê and î in Kurdish, ā in Latvian, and so on (it seems like my sorting script has ripped apart some Vietnamese characters...). It also conflicts with Ags not sending key codes for the shift keys. It will be possible to find a compromise, I guess, once I have received my AG and know which characters are available for the new layout.

Have the rarely used characters on the front

That is for not having to memorize them.

Minimize hand switching while typing

Since I was seriously planning on creating a keyboard similar to the AG for about a year I have been watching out for typing mistakes. There are tree types:

a) A missing letter

b) A letter to much

c) The wrong letter

d) permutations

There is not much to do against type (a). Type (b) and (c) can be improved by an improved layout; the asymmetric standard keyboard layout really asks for landing in between or on the wrong key. Type (d) makes up more than 90% of all typing mistakes (that’s an educated guess, I didn’t keep a list). They always (100%) occur when switching hands, never when both letters are on the same hand! That’s why Dvorak turns out to be a very bad idea – it maximizes typing errors. A user-friendly layout would try to minimize hand-switches. Sadly character combinations are somewhat language dependent, even though enforcing (1) does not make matters too bad.

Replace {Print}, {Scroll lock}, {Pause} and {Caps lock} and possibly the duplicates of the {Shift}-keys by something more useful.

Will anyone die without those? If one Shift key is moved to the back side, there is no need for a second one. Sadly this is not possible if it doesn’t send a key code.

I would welcome comments on the conflicts noted above, especially: How important is 4a? Also: How should the languages be weighted?

It would be absolutely great if someone could make a list of key codes (e.g. using xev). I would like to put my results, character counts, and thoughts online somewhere. I checked, but my university doesn’t give server space for student home pages, so can anybody recommend a place where I can put up a web page (best for free without too annoying ads)?

Try to let both Hands do an equal amount of work

This is to avoid one hand hurting after typing for a long time, but it heavily collides with (3). You can have either few hand-changes or do an equal amount of typing left and right. The result when enforcing this point still won’t be as bad for typing mistakes as QWERTY and still much better than Dvorak, but won’t be much of an improvement either.

Try to allow as much one-handed usage as possible

It would be very convenient to be able to turn a page or pick your nose without slowing down in typing. Together with (1) this will hardly be possible to archive for normal text, but it is worth thinking about making the “Number Pad” one handed, so you need only one hand to enter numbers and keep the other hand for pointing on the line at the paper you are copying of. Also one hand could be responsible for typing while the other hand does navigating (ctrl, Arrows, Page ↑/↓, Home, End, Tab,...), which is the direct opposite goal to 4a. This demand – surprise – coexists perfectly with (3).

Additional Commentary

From: Matthias Schult Date: Sat, Apr 29 2006 8:59 am

On Sat, 29 Apr 2006 06:06:25 -0000 "Didgers" wrote:

> 1) I looked at the various language specific Dvorak layouts, and most

> looked like the American original.

> You could start by getting a hybrid

> out of them. The philosophy of Dvorak Layout is to reduce the effort

> required on a traditional keyboard.

Sorry, it seems not to have become clear what I intended to do. I want to create a layout based on scientific research as much as possible. The reason is not only to get as close to the optimum as possible, but also to keep 1001 different layouts popping up everywhere. Starting with a layout optimized for a standard keyboard thus is not the way to go. The way to go is creating one from scratch.

Anyway, as an example I compared German and American Dvorak layout. They differ in 25.9% of an average German text and in 35.0% of an average English text -- even though both languages are Germanic languages. I wouldn't call that very similar...

The Dvorak layout is based on:

1. minimizing finger mileage: This does not apply to the AG since most keys are right under the fingers anyway
2. maximizing hand switches: Dvorak thought that would be faster, which I strongly doubt. There have been no serious scientific investigations about that which I am aware of. The only study I know shows a tendency that typing with adjacent fingers inwards (ring, middle, index finger in that order) is the fastest combination possible, but even that might be different if not typing on a flat board. On the other hand, hand switches are very error prone.

> They may not have been thoroughly

> optimized for their language, but I'll bet it's better than what was

> derived from qwerty. Anyhow, there's only so many keys on the AG-5,

> but there are so many international characters. It would be difficult

> to get them all on there, and still retain a level of ease of use.

> AutoHotkey can only do so much.

That's a misunderstanding. I mean you don't have to be able to reach all international characters on every layout, just the language specific ones. But those letters should be the only serious difference between the layouts. I don't have MS Windows, so I don't need AutoHotkey, but I'm confident that the resulting layout will be possible to implement in it.

> 2) Since different languages use different letters, some letters'

> rarity will vary. Assuming you can even get all of the characters on

> the AG-5, there would be at least one language that will have a steady

> stream of front pecks.

No, surprisingly. The 10 most common characters make up between 60% and 65% percent of a language for almost every language. And those include a,e,i,n,r,t and space for nearly all languages. The order of the 20 most frequent characters might be different from language to language, but the characters are basically the same. Things start getting different when looking at character combinations and digraphs.

> Probably ought to just settle for a layout

> that's good for two or three different languages or good for a family

> (Germanic, Romantic, Cyrillic, etc)

Does a romantic layout have l,o,v,e in the home row? :-) I won't make any Cyrillic layouts, because I don't speak any languages which use the Cyrillic alphabet myself. So I stick to the Latin alphabet. The 5 language families with the most speakers in languages using the latin alphabet are:

• italic (French, Spanish, Italian, Portuguese, Romanian, ...)
• germanic (English, German, Dutch, Swedish, Norwegian, Danish, Luxemburgish, Islandic, Afrikaans, ...)
• malayo-polynesian (Indonesian, Tagalog, Javanese, Cebuano, ...)
• turkic (Turkish, Tartar, Azerbaijani)
• slavic (Polish, Czech, Slovac, Slovene, Serbo-Croatian, ...)
• Though small, Finno-Ugric (Finnish, Hungarian, Estonian) is pretty far spread in the internet as well.

> 3) My only mistakes are missed / wrong keys. I don’t really have too

> much trouble with permutations, but that might just be me. I honestly

> believe that Dvorak’s alternating hands rthym helps against permutation

> error.

My observations are mainly based on e-mails I got, less on my own writing. It’s also a conclusion I didn’t expect. I expected missed or wrong keys to make up most of the typing error, but they clearly don’t.

> 4) If you want to type with one hand, you could try to base your AG-5’s

> layout on one of Dvorak’s one handed typing layouts. Whatever layout,

> you will be overwhelming the physical limits, which are already being

> pushed due to all the international characters you’re trying to type.

No, it’s not about one-handed writing. My formulation probably wasn’t so good. It’s about what’s more convenient for normal use.

I would reserve chording for stenographic typing, except maybe for writing language specific stuff. E.g. in German umlauts are resolved to ä→ae, ö→oe, ü→ue, so those chords would be kind of natural. …