Japanese Typewriters

> It all sounds horribly complicated, although it may be one of those procedures that's easier to do than to describe.

Definitely. It's still annoying when you type half width in a field that requires full width or something like that.

But the keyboard changing feels natural - only badly designed UIs feel clunky.

There has been the stylus tech for a while to do good input, but typing out phonetically is basically much quicker (and the "flick"-based inputs of touchscreens even faster). Do you want to write out like 7 lines on a screen for a character? or just type its 2-4 letter phonetic?

The current input stuff is basically like T9 crossed with gmail's autocomplete: it's really good at predicting what you want for the most part, and lets you type at very fast speeds whne you're good at stuff. Exaggerating a bit with the GMail autocomplete but some very common fragments will pop up almost instantly when you start typing them in.

In a sense the input methods are a solved problem.

> It all sounds horribly complicated, although it may be one of those procedures that's easier to do than to describe.

It's extremely annoying to be honest.

I fully believe that logographic scripts are faster to read once properly mastered, but writing or inputting them is another matter altogether and the article doesn't even begin to touch upon some of the frustrations such as the i.m.e.'s intelligence deciding at the last character to change the educated guess, completely changing the sentence from what intended as one presses “enter” or the inability to type without looking at what one types which is quite possible in many other scripts.

The system tries to be as intelligent as it can and learns from patterns so the guesses are usually correct, but it also means that at times it is unprædictable.

It can also be quite disorienting for a sentence to suddenly shrink or expand considerably as one is typing, and it isn't generally easy to see part of a word one has typed; only when the full word be typed is the proper conversion made and before that point the i.m.e. actually converts it to something entirely differently, often completely misinterpreting the structure of the sentence because a sentence that ends on an incomplete verb is not a sentence grammatically speaking so the i.m.e. doesn't know what to do with it.

Then comes the fact that the i.m.e. is generally very bad with colloquial Japanese as it's conversion rules rely on the grammar of standard textbook Japanese.

Having said that, there is an odd sense of satisfaction to typing Japanese when the i.m.e. agree with the intent of the user. There is something satisfying about a garbled sentence that looks like nothing instantly transforming into a coherent sentence as the last character is added to it.

> I wonder if things will change [or have changed] with the advent of touch screens and styluses? [stylii?] which would allow the characters to be written manually, using finger or stylus and then have the software convert this to the appropriate typed character?

Yes, the system is quite a bit different there; as the article says: Japanese characters are primarily arranged in a grid of with a width of 5, so te standard approach there is to drag in one of four directions, or not drag at all to select a character, and a list of possible conversions to Chinese characters is more easily available with a touch screen to select from. In this sense typing Japanese on a mobile phone is actually not any more time consuming than typing English on a mobile phone, but on a keyboard, there are definite problems with it.

Technically it's incredibly complicated, unreliable and crazy it even works at all. The upside is that mobile phones have the same problem space, and Japanese was ready from get go to deal with autocomplete on crammed touchscreens.

From a user point of view, you get used to it and it's not so bad after a while. The irritating parts (e.g. having to get back to the middle of your sentence to fix a typo otherwise the surrounding characters are also messed up) are offset by the ease to input emoji, pictograms, or basically any unicode char. you remember the name is a godsend ♡

I don't think Japanese IME is difficult. You just type the word as it is pronounced.

Typing on a smartphone is more difficult because the keyboard has only 10 keys for characters and you have to tap a key and slide up/down/left/right to chose one of 4 characters.

For typing in keyboard, typing Japanese is decent speed in most cases even though IME isn't perfect. It's a bit pain combined with Vim since both IME and Vim has mode.

For input in smartphone, flick input is great way to input but we miss Swype style (or say predicting style) input in Japanese Qwerty keyboard.

Writing Japanese on screen isn't great since Japanese forgetting writing kanjis.

It's very straightforward, and although typing speed is not quite the same as English (for me, at least, although I am not a native speaker), it's not wildly different.

Typing a Japanese word or phrase is just like typing in English, with the exception of selecting the appropriate kanji from a list (no mouse or stylus required). An experienced typist can do this very quickly, on a physical keyboard or on a smartphone.

For a word like "Japanese" (language), i.e., Nihongo or 日本語, it's actually fewer keystrokes than in English: 5 vs. 8, or 9 if you count the space afterward since Japanese doesn't have the same spacing requirements.

I was deeply impressed when I visited Japan in early 2000's and saw people with flip-phones typing in Japanese at speed comparable to what I saw people do on a flip-phone using English. It looked like they had auto-complete available on phones a decade before we saw it in the US!

I've been studying Japanese for about 10 years now and think these systems are actually rather intuitive if you know the language. Digital input methods aside, most of the complexity comes from the writing system itself and just comes with the territory.

Phonetic IMEs aren't the only input methods available. There are a number of input methods based off of character components for Chinese and Japanese that are generally an order of magnitude faster (not accounting for predictive text). It requires about the same level of memorization as learning Vim. The problem with phonetic input is that it doesn't narrow down the search space enough for typing to be open loop. It's essentially the equivalent of typing in IPA and then selecting the equivalent English spelling. The problem is that without predictive text the search is too broad for any given input. Kanji actually has a set of about ~500 atomic components that all characters can decompose down to, so if you just map these in a lossy way to the US keyboard or just the numpad for some input methods, the search space is narrow enough that you can get close to open loop typing.

You can also decompose to about 10 strokes, whose order when written is set, so you can map from sequences of those strokes.

The problem with all these input methods IMO is that they're not deterministic and therefore not open loop because that just wasn't one of the design goals.

The additional benefit is that you can apply whatever rules your input method uses to symbols from math or other languages, which IMO is far simpler than using LaTeX syntax if you just want to type a ‰ or ⊥ sign. My input method can type Chinese, Japanese, Greek, and Russian as well as some mischellaneous symbols.

Entirely mechanical component based typewriters were being developed in the 50s but never came to fruition. By the time electronic typewriters came along it was decided that it was just cheaper to use the existing physical keyboards designed for Ascii or numpads on phones and work around it with software. This seems to have been a consistent pattern for most languages regardless of the number of symbols or whether they were phonetic. IMO I believe if they had just standardized on a more specialized keyboard layout in the 80s IMEs would not exist. IMO you can likely deterministically type Kanji in a open loop way if there were maybe 2 to 5 times the number of physical keys as the standard, or the same number of physical keys with 4 more shift keys. The advent of predictive text the need for it is basically nil for the majority of people in the same way the majority of people have no need for LaTeX or Vim.

Kanji, like many Asian scripts, has no word spacing. So what is and is not a word is a bit of a hard problem. Implementing word count for kanji is a non trivial problem. There's a Norwegian that did a machine learned stemmer for Kanji for Elasticsearch at some point to group characters in meaningful groups.

Probably characters per minute is a better comparison but it's still apples and oranges as you typically need less characters in kanji to spell out the same thing.

If you follow the link there's fullscreen hardcore pornography playing behind the transparent age warning. Incredibly not safe for work.

And that's why uBlock Origin is your best friend. No ad for me was shown while uBO count was 20+ for blocked elements.

ಠ_ಠ