Naginata Style (薙刀式), by author Toshihiko Ōoka (大岡俊彦), is a Japanese kana-based keyboard layout which caught my interest. It was developed to work on traditional keyboards with spacebar used as central shift, but also works well on small ergonomic keyboards with thumbs-keys. The most common kana are a single key press, but generalising the idea of shift keys any other kana requiries a two-key combo (including compounds like "きゃ" or "kya"), or at most a three key combo (compounds like "ぎゃ" or "gya" with a ten-ten modifier). The layout is more ergonomic than Qwerty and the rarely used JIS Kana layout, and needs less key presses than romaji.
The bulk of this post is my current understanding of the core layout, and some of the design principles behind it. This is based on trying the QMK and ZMK implementations, and computer translations of the documentation and blog posts. I am likely missing some concepts. For context, I've passed the easiest JLPT Level 5 exam, and with a bit more kanji practice ought to pass Level 4.
This layout was first released in January 2018 on the Naginata Style (薙刀式) blog post, and has been continuously refined since (old versions at the bottom). As of April 2025, it is on Naginata Style v16 (that blog post describes the changes). As the extensive manual notes, this is more than just a layout, as it defines rich editing layers and shortcuts as well.
The creator Toshihiko Ōoka (大岡俊彦) is an author, so he writes a lot. He also blogs a lot, including loads of content about keyboards. Things like videos of various Japanese keyboard systems (May 2020), was no support for Japanese thumb-shift keyboards in Windows 95 a factor in their demise (Augut 2025), and Japanese entry innovations since the thumb-shift in the 1980s (August 2025). Some posts are easier to understand with computer translation that others. This one talked about the fraction of typed kana needing a single key on various layouts, and included this figure showing the "50 sounds" scaled by usage:
You can see how those most used kana get prioritized in his layout below.
Naginata Style
The layout makes heavy used of chording or combos - pressing one, two, or three keys at once.
The layout can be used on a traditional keyboard (with remapping software), or a custom programable keyboard. It requires just 31 keys, a block of three of five for each hand, and a shared central spacebar/shift thumb-key. Enter is available as a combo (Qwerty "V" and "M"), but is sometimes given key 32 as another thumb key on an ergonomic keyboard layout (a combined tap for enter, hold for central-shift thumb-key).
Backspace gets the right hand top-row index finger position (Qwerty "U"), contrasting the traditional placement top right outside the core block which is extremely poor ergonomically as the entire hand must be moved.
Basic kana
There are only 25 kana plus the long-sound 「ー」 on the base layer, which therefore only need a single key press. Perhaps surprisingly this includes just three vowels, laid out as the right hand home row of 「あいうー」 ("a", "i", "u" and the katakana long-sound). The less used vowels 「お」 and 「え」 ("e" and "o") are demoted to shifted-keys (still on the right hand). In general my understanding is the more common kana are allocated to the stronger index and ring fingers. The final four keys on the base letter are backspace (discussed above), 「小」 (meaning small, mainly for the small vowel forms), and back/left and forward/right cursors at Qwerty "T" and "Y" (for editing and kanji selection). The shifted layer has the rest of the basic kana, 「。」 (Japanese full stop) and 「、」 (Japanese comma), and a few spare slots (not enough for all the standard punctuation and navigation controls which require alternative layers, see below).
Ten-ten and Maru modifiers
Taking an idea from the earlier Japanese thumb-shift layouts, the ten-ten modifiers (濁音 dakuten) are combos with other hand's index-finger on the home row (Qwerty home keys "F" and "J", denoted 「濁」here). For example, 「か」 (ka) is on Qwerty "G", and when combined with 「濁」(Qwerty "J") gives 「が」 (ga). Either key can be held down like shift while the other is tapped, but as long as both are pressed and down at the same time this suffices making it easier to do that it might sound. The exact typing feeling will depend on the implementation.
Likewise, the less common maru modifiers (半濁音 handakuten) are combos with the other hand's index-finger on the bottom row (Qwerty "B" and "M", denoted 「半」 here). For example 「ひ」 (hi) on Qwerty "X" when combined with 「濁」 (Qwerty "G") gives 「び」 (bi), but combined with 「半」(Qwerty "M") gives 「ぴ」 (pi).
Importantly, where there are two kana sharing a key with shift, only one has a ten-ten or maru form, meaning shift can be omitted leaving a 2-key combo. Note foreign sound 「ゔ」 is an exception, see below.
The 「濁」(Qwerty "F" and "J" for ten-ten, they are also kana 「か」 and 「あ」) and 「半」(Qwerty "B" and "M" for maru, they are also kana 「こ」 and 「な」) keys are highlighted as pale green squares on the layout picture below:
Small kana
Most small kana are a 2-key combo using the Qwerty "Q" key labeled 「小」 on the left hand in combination with the key for any kana with a small form without needing shift. That covers all of あいうえおやゆよわ」 (a/i/u/e/o/ya/yu/yo/wa) on the right hand. This also includes rarely used cases of small "ka" and "ke" (shown as katakana only, 「ヵ」 and 「ヶ」), which are a little awkward as they are also on the left hand.
The exception is small 「っ」 (xtsu) which is only available via its own key (Qwerty G). Qwerty L is used for 「う」 (u) which has a small form 「ぅ」 (xu), and large 「つ」 (tsu) when shifted. Using 「小」+「う/つ」gives small 「ぅ」.
Small 「ゃゅょ」 (xya/xyu/xyo) in a compound kana like 「きゃ」 ("kya") are instead normally entered as a two or three key combo, discussed next.
Compound kana
Two key combos between any of 「きしちにひみり」 (ki/shi/chi/ni/hi/mi/ri, all on the left hand, some usually needing shift) and one of 「やゆよ」 (ya/yu/yo, all on the right hand and requiring shift when used alone - written in red on the layout picture above) give the compound kana (拗音 , yōon) like 「きゃ」 ("kya") without needing shift.
These become 3-key combos when including a ten-ten or maru modifier, like 「ぎゃ」 ("gya") or 「ぴょ」 ("pyo").
Foreign sounds
The same approach is used for the foreign sound compound kana, usually written in katakana. They are written as 3-key combos of a consonant, either 「濁」or 「半」, and a vowel (no need for shift).
For example, the katakana 「ティ」 would be 「てぃ」in hiragana ("texi" or "thi" in romaji), and is typed with て+半+い. Similarly, the katakana 「ディ」would be 「デぃ」 in hiragana ("dexi" or "dhi"), and is typed with て+濁+い.
The rarely used 「ゔ」 ("vu") is an exception. As noted above, Qwerty L is used for 「う」 ("u"), and large 「つ」 ("tsu") when shifted - which both have small forms. These also both have dakuten modified forms. Qwerty F 「濁 」and L is interpreted as modifying (implicitly shifted) 「つ 」 giving 「づ 」. Instead 「ゔ」gets a three key combo with the right hand also pressing Qwerty semi-colon 「ー 」(the neighbouring key). Some earlier revisions to the Naginata Style used Qwerty Q for 「ゔ」 ("vu"), which is now used for small kana.
Elegant Core
I think of the chording or combos at the heart of Naginata Style as generalising shift modifiers:
- Central-shift (the space bar), required with nearly 50 kana to fit in under 30 keys,
- Ten-ten-shift 「濁」(the Qwerty F and J keys), for the 濁音 or dakuten,
- Maru-shift 「半」(the Qwerty V and M keys), for the 半濁音 or handakuten,
- Small-shift 「小」(the Qwerty Q key).
By their nature for any kana, the ten-ten, maru, and small "shift" are mutually exclusive. The genius of the layout ensures only one of these four "shifts" needs to be used at a time - the central-shift can be ommitted and inferred. To do this the layout ensures for each key with two kana (the default and centre-shift alternative), only one has a dakuten or handaukten form, and only one has a small form. That means all the single kana need just a single key press, or a two-key combo.
[There is the exception discussed of Qwerty L being both 「う」 (u) and 「つ」 (tsu) when shifted, which both have small forms and dakuten forms. I suspect this might be tweaked again in a future revision.]
The compound kana like 「しゃ」 (sha) consisting of a large kana in the い-row (on the left hand), and a small kana from 「あいうえおやゆよ」 (a/i/u/e/o/ya/yu/yo on the right hand) also get combos - the key for the large kana and the small kana respectively (where again the centeral-shift can be ommitted and inferred), and if required the (han)dakuten shift as well. This means all the compound kana are two or three-key combos.
Layers
Although the base layout manages to squeeze in left and right cursors, for all the other navigation keys you have to resort to layers on a minimal keyboard. Here Naginata Style defines two-key combos of the ring and middle finger on one hand activate a layer for the other hand.
Holding Qwerty "D" + "F" on the home row with the left hand gives vertical movement on the right (and more), while Qwerty "C" + "V" on the bottom row with the left hand gives horizontal movement on the right plus editing shortcuts (copy, cut, paste, undo, redo) etc.
Holding Qwerty "J" + "K" on the home row with the right hand gives assorted punctuation including quote marks and brackets on the left hand, while holding Qwerty "M" + comma on bottom row on the right gives more bracket pairs and other functionality on the left hand.
From skimming the computer translation, most of these extra functions are related to text editing. The clear separation of horizontal navigation and vertical navigation probably stems from how the author conceptualises writing Japanese vertically (the more traditional style in columns top-to-bottom and then right-to-left), or horizontally (as in European languages). This is something which changed in v16, with earlier versions explicitly configured for either vertical or horizonal writing. The author suggests modifing this to suit - for myself I have an inverted-T set of cursor keys for the right hand activated via a thumb-key.
Interestingly there isn't any obvious way to type the digits 0 to 9, which I presume is done by swapping out of Japanese mode and back to alphanumeric. I suspect the author primarily writes numbers with their kanji 「一二三四五六七八九十百千万…」 instead.
Implementations
It appears the original version was implemented in software on Windows using something called DvorakJ (last updated in 2014), but the main Naginata Style page lists multiple other volunteer maintained implementations - mostly by Eswai. Of these, I have currently successfully built and tried their QMK Naginata (currently v15) and ZMK Naginata (currently v16) for my own programable keyboard.
On macOS, there is a version of Naginata for Karabiner-Elements (currently v14) which has multiple forks including some efforts to update this, which I may try to integrate into my own Karabiner-Elements laptop ergonomic layout (used for English).
Simple Example
To type 「私も名前はピーターです。」 for "My name is Peter." in romaji would be "watas(h)i no namae ha pi-ta- desu." (spaces for clarity, 26 or 27 keys plus enter). Using Naginata Style:
- Centre shift + Qwerty "." for "wa", 「わ」
- Qwerty "N" for "ta", 「た」
- Qwerty "R" for "shi", 「し」 (at this point the IME made 「私」)
- Centre shift + Qwerty "J" for "no", 「の」
- Qwerty "M" for "na", 「な」
- Centre shift + Qwerty "F" for "ma", 「ま」
- Center shift + Qwerty "O" for "e", 「え」 (at this point the IME made 「名前」)
- Qwerty "C" for "ha", 「は」
- Qwerty "X" + "N" for "pi", 「ぴ」 - optionally with shift for "PI" for 「ピ」
- Qwerty ";" for "-", 「ー」
- Qwerty "N" for "ta", 「た」 - optionally with shift for "PI" for 「タ」
- Qwerty ";" for "-", 「ー」 (at this point the IME made this into katakana「ピーター」automatically)
- Center shift + Qwerty "E" for "de", 「で」
- Qwerty "O" for "su", 「す」
- Centre shift + Qwerty "N" for ".", 「。」
- Qwerty "V" + "M" or Enter to accept the IME actions (all correct in my case).
That's 15 single key presses or combos, plus enter. Even counting key presses, slightly less than romaji.
Kanji Example
Consider this slightly contrived example to force using the IME kanji selection: As romaji, "hashi wo motte hashi wo wataru." (with spaces added between words for readability). In hiragana, 「はしをもってはしをわたる。」 where 「はし」 can be 「橋」for bridge, or「箸」for chopsticks. The expected choice of kanji is 「箸を持って橋を渡る。」and this translates as "Cross the bridge carrying chopsticks". The macOS IME will also offer the emoji 🌉 and 🥢, as well as these and other kanji.
As an aside, note the "small tsu" 「っ」 gets its own key (Qwerty "G"), distinct from the "large tsu" 「つ」 (Centre shift + Qwerty "L"), which is quite unlike the repeated consonant approach in romaji.
How does kanji selection work? Just like it would in ordinary romaji mode. Specifically in macOS (where I have only used horizontal left-to-right Japanese writing), you can adjust the suggestions as you go along, or more typically at the end of the sentence. Here you would typically use the left/right cursors to move to a wrongly guessed block (easy here with those on the index fingers; other shortcuts exist like readline style ctrl+b and ctrl+f), and then change the suggestion from the list using either space or down for the next suggestion (common), or shift+space or up for the previous suggestion (rare unless you've pressed down too much). Finally when all is well, press enter to confirm the suggestions, and resume typing. Escape can be used to undo an automatic kanji conversion.
Selecting the alternative kanji suggestions puzzled me. We don't have up/down arrows on the base layer (although in earlier iterations of Naginata Style, the layout can be setup with up/down instead of left/right, primarily for when writing Japanese vertically I believe). We do have the central spacebar available - the Naginata implementations I have tried make this behave as a tap-or-hold key: Tap for a space (useful for kanji selection), hold for the Naginata central-shift.
So with a minimal keyboard with just the 31 essential keys, how do you select the previous kanji in the list (or cancel a conversion)? You can't type shift-space (or escape) on the base layer. The Japanese manual is hard to follow with computer translation, but I think you have to use an editing layer for the missing cursor keys (see above).
Or, use more keys - most programmable ergonomic keyboards have at least a few more - so you might have a traditional left and right shift. That can be used for typing shift+space in kanji selection, or to make the romaji appear in upper case, which on the default macOS Japanese IME can be configured to be interpreted immediately as katakana. I'm trying out a thumb key giving shift-space when tapped (for previous kanji selection), and traditional shift when held (for katakana).
Or, you can define your own layers with cursors (which I do too, see above).
What next?
I need to find more reasons to type in Japanese, and practice my kanji too. Perhaps with Wani Kani on the computer, but I tend to use that on my phone. I'm hoping a kana-based layout will help me stop thinking in terms of romaji (flick input on a phone helps there too).
No comments:
Post a Comment