Kana-Chording-KE : Flick-input like Japanese keyboard chording

Concept

By keyboard chording we mean pressing multiple keys at once to generate characters. This idea has been applied to Japanese kana which are laided out on a 5×10 grid (the gojūon, 五十音, "Fifty Sounds"), starting with the five vowels (a, i, u, e, o, or あ, い, う, え, お), followed by the ka-row (ka, ki, ku, ke, ko, or か, き, く, け, こ), then the sa-row, etc. We can address any of the matrix positions with a consonant row (a, ka, sa, ta, na, ha, ma, ya, ra, wa) and vowel column. Thus combining ha and i gives hi, は＋い→ひ).

あぁ	いぃ	うぅ	えぇ	おぉ
かが	きぎ	くぐ	けげ	こご
さざ	しじ	すず	せぜ	そぞ
ただ	ちぢ	つづ	てで	とど
な	に	ぬ	ね	の
はばぱ	ひびぴ	ふぶぷ	へべぺ	ほぼぽ
ま	み	む	め	も
ら	り	る	れ	ろ
やゃ		ゆゅ		よょ
わ		ん		を

Plus っ (xtu), ゔ (vu), ヵ (xka), and ヶ (xke).

This is a software based project inspired by:

• KanaCord which was a hardware project building a physical 30 key chording kana keyboard with internal software. This was a split keyboard where the left hand had 2 rows of 6 keys containing maru (ﾟ, handakuten), ten-ten (ﾞ, dakuten), and a small key (labelled 大), making 3 modifiers, and the main consonants (9 keys). On the right were 3 rows of 6 keys, containing the vowels (5 keys), ん, space, assorted punctuation, and a hiragana/katakana toggle. The core idea is the (almost) 50 main kana are typed by pressing the appropriate two keys together.

• The 12-key iOS Japanese flick keyboard for touch screens, which lays out 10 keys for each row (a, ka, ..., wa) plus context dependent size toggle, maru, or ten-ten modifier, and a core punctuation key. Here each key can be tapped giving the labelled character from the a-column, flicked left (i-column), up (u-column), right (e-column), or down (o-column).

This project combines these ideas whereby any of the ASCII keys representing a row (like k for ka, ki, ku, ke, ko) can be pressed alone or in combination with one of four modifier keys (initially the four cursor keys) giving the five different vowel combinations.

Implementation

We are using the very capable macOS keyboard modifying software Karabiner Elements to do this on macOS. KE has its own rather verbose JSON configuration language which can be used to define complex reboard remappings, including multiple keys at once which we need for chording. Creating over 100 JSON rules for all the kana had to be scripted, for which I am using Python.

The KanaCord uses operating system specific keyboard expressions to enter the hex code for individual unicode for each kana. For example, け (ke) is unicode 3051 in hex.

For macOS, once in "Unicode Hex Input" mode, press-and-hold the option key (aka alt), then type 3, 0, 5, 1, and release the option key. The top number row or the number pad can be used. This will insert unicode hex character 3051, け (ke). The same trick works for emoji too, usually eight digits, and can be done from KE. Unfortunately, using "Unicode Hex Input" mode has the significant drawback that you won't get access to the OS' kanji support. In this mode hiragana and katanana would both work by directly sending the unicode hex for each character.

If instead we use "Japanese - Romaji" mode, we must map the desired kana to any of the supported romaji aliases, and send that key sequence. Thus if the user presses right+k wanting け we map this to sending k then e, and let the OS turn this into the kana or kanji as more is typed. We can also support shift+right+k sending upper case K then E which can be configured in the macOS Japanese settings to be treated automatically as the katakana ケ. In this mode to enter coffee we would want to simulate pressing K O - H I - giving コーヒー as you typed. Here is a great introduction to typing in romaji mode which covers this and more.

Alternatively, in "Japanese - Kana" mode, we must map the desired kana to the physical key on the macOS Japanese keyboard, and send that keycode. In KE these are labelled by the English letter or symbol on that key (these keyboards have a full QWERTY layout with USA like punctuation placement). Thus if the user presses right+k wanting け this corresponds to the semi-colon key : so we send that. When right+g is pressed for げ, we must also send the ten-ten key @ (and likewise for maru modifications follow the base kana with the maru key [). In this mode the shift key gives the small form of the vowels a/i/u/e/o, tsu or ya/yu/yo, and toggles the わ key to を (key zero). For katakana, in this mode to enter coffee we would want to simulate pressing b ¥ v ¥ giving コーヒー via the language support. There are explict keys for ◌゙ (adding ten-ten or dakuten) and ◌゚ (adding maru or handakuten).

Physical Keyboards

Full querty

The initial mapping is based on the flick-input on iOS using the cursor keys. Thus k alone gives ka (か), while left+k gives ki (き), right+k gives ke (け), etc. This can be used with the QWERTY layout on a Japanese keyboard, or anything similar with "Roman" letter like any English layout, or the French AZERTY for example.

For example, n+left (→に), h+down (→ほ), w+up (→ん), g+down (→ご) would type にほんご which should offer the kanji 日本語 (meaning Japanese language). That is four chords (three and a single character if you prefer), compared to at least seven keys in Romaji mode (either nihongo or nihonngo works).

For an example with ten-ten and maru modifiers, the Japanese word for grammar is ぶんぽう or 文法. This can be entered as b+up (→ぶ), w+up (→ん), p+down (→ぽ), a+up (→う), four chords for four characters.

For a katakana example, コーヒー (meaning coffee) in romaji mode using shift to indicate katakana explicitly, would be shift+h+down (→コ), - (→ー), shift+h+left (→ヒ), - (→ー). That is four chords (two and two single keys if you prefer), compared to six keys with shift in Romaji mode (KO-HI-).

This may prove popular with Japanese learners, as no new key locations need be learnt, only a couple of corner cases like up+w→ん to match the iOS flick layout. However, with the right hand on the cursors to mimic the flick action, the left hand must cover at least the 10 keys a, k, s, t, n, m, y, r, and w (plus optionally the 5 ten-ten and maru modified forms using g, z, d, b and p, and 2 small forms using l and x) which on QWERTY layouts are split between the left and right halves of the keyboard making it impractical. An alternative mapping taking only keys from the left hand side would be easier physically, but harder to memorise.

12 key grid

Instead my initial plan is to use a custom keyboard mapping on a supplementary small macro keyboard with at least 12 keys, mimicking the iOS flick-input with its core four rows of three (additional controls either side not shown):

あ	か	さ
た	な	は
ま	や	ら
大	わ	、

How to handle the context dependent ten-ten, maru, or small form kana modifier (shown here as 大) remains to be decided (three modifier keys woud be simplest), while for the punctuation button the English comma makes sense:

a	k	s
t	n	h
m	y	r
?	w	,

The earlier example using core hiragana remains the same: n+left (→に), h+down (→ほ), w+up (→ん), g+down (→ご) would type にほんご which should offer the kanji 日本語 (meaning Japanese language).

Words with ten-ten and maru modifiers like ぶんぽう (文法) would require some kind of shift/modifier. As would small characters.

For a katakana example, コーヒー (meaning coffee) assuming we don't have access to the hypen key -, we can use an iOS flick-input inspired chord comma+down (→ー). Thus four chords shift+h+down (→コ), comma+down (→ー),　shift+h+left (→ヒ), comma+down (→ー) for four characters. (Perhaps we should make both comma+down and shift+comma+down → ー since this will often be used with romaji-shift mode?)

Additional keys would be needed for kanji selection, proably at least escape to cancel, space to cycle the suggestions, and enter to accept. Perhaps for the right hand?

15 key grid

A variation of this idea would be to use a custom layout on a macro keyboard of at least 15 keys to replicate the Gboard Godan キーボード flick-keyboard which has separate keys for the five vowels (a, i, u, e, o) as the left-most column, and the nine consonant rows (k, s, t, n, m, y, r, and w) as the central and right-most columns. These are shown as upper case Roman letters (with additional controls either side not shown):

A	K	H
I	S	M
U	T	Y
E	N	R
O	小	W

Again the ? here is a placeholder for a context dependent modifier.

My understanding is to type 日本語 would need seven keys, just like Romaji mode: N I (→に) H O (→ほ) N (→ん) G O (→ご).

In this approach the flick actions left used for ten-ten (like H to B), right for maru (H to P), down for the numbers (telephone keypad layout), and up for corner cases. This is also used in ABC mode to access the rest of the alphabet, meaning in principle this might replace a physical QWERTY keyboard rather than just supplementing one - although you'd still need cursor controls, escape, enter, etc. The middle key in the final row is context dependant for toggling the modifiers (redundant?) or size:

・ ・ A ・ 1	Q ・ K G 2	F P H B 3
・ ・ I ・ 4	J ・ S Z 5	L ・ M ー 6
・ ・ U ・ 7	C ・ T D 8	X ・ Y ・ 9
・ ・ E ・ ・	・ ・ N ・ 0	？ . R ！ 、
・ ・ O ・ ・	◌゙`   ◌゚ 大 ⇔ 小	V ・ W ・ ・

I don't see how this might be used with macOS "Japanese - Kana" mode, unless when a consonant is pressed we wait for the vowel before sending anything to the keyboard? However, it seems a natural fit to the "Japanese - Romaji" mode.

In this layout, for typing Japanese the only flick actions (here chording of some sort) would be for the ten-ten and maru modifiers - leaving the second hand mostly idle, perhaps mainly handling the kanji selection? Rather it seems logical to split the layout as per the intended thumb split - one hand gets the five vowels, and the other the nine consonants - taking us back to the KanaChord keyboard approach.

License

MIT License