Typing frustrations

So I’m on Windows 10, I have downloaded the Japanese IME, and type on my qwerty keyboard what is essentially romaji, which the system first turns into hiragana, then suggests kanji (or even katakana) for what I typed. Most of the time this works like a charm, but there are times when the system just can’t guess what I’m writing, and won’t suggest the kanji I need. For now I get around this by typing a different word that uses the kanji I have in mind, then deleting any extra hiragana or kanji the system adds that I don’t need for what I’m typing. This seems rather impractical to me though, so my question is, is there a better way?

In the same vein, I’ve always wondered how people in Japan type on their computers. I’m sure there are Japanese keyboards so they probably type straight hiragana instead of romaji, but how do they handle the kanji? Do they again let the system decide? And what if the system isn’t being helpful?


I’m not an expert, but do you have an example where the system doesn’t provide you with the proper kanji? It might be easier to give tips if we have something more concrete to work with.


Are you trying to type full words or individual kanji? It’s rare that my IME can’t figure out how to convert a full word to its kanji form.


It happens but should be pretty rare. Usually if you add more context (i.e. type a longer phrase or at least add a particle) the correct kanji should pop up.

Btw, I use Google IME on Windows because I think its suggestions are better.


I know, I’ve tried to reproduce it, but it seems like when I’ve successfully managed to enter a word, the next time it’s the first suggestion. 注射 was one such word, strangely. I had to write it part by part before, now it’s number one when I type ちゅうしゃ.

Full words always, and often I type the whole sentence before I “commit” the spelling. I was thinking that maybe that might be the problem. Sometimes more context seems to help, and sometimes it just seems to confuse the system. Do you commit the correct spelling as soon as it pops up, or do you finish typing first? I don’t know what would be best practice.

1 Like

That’s because the IME learns which words you are using. Over time it will become more helpful.

I prefer to confirm input in short phrases.

If the word is not common then sometimes I have to go through multiple pages of suggestions, which is why I prefer working with a short phrase. But overall I don’t think I’m struggling too much with my IME.


Yes, I can see that the more I type the easier it will get. I really wonder how people typed in Japanese before such “smart” systems existed. Although I wish it wasn’t always so smart. I dislike it when it suggests context I never typed (adding particles for example), and ends up having the kanji I need (with no additions) on the next page of suggestions.

1 Like

Before such smart systems existed (seems like they started to be invented around the 70s), normal people didn’t have much cause to need to type. I mean, no personal computers or email, either.


Smart predictive systems that learn from your behaviour are very recent though. Typing has existed way before computers. There were typewriters (that must have been fun in Japanese), and there was definitely typing on computers back when even a simple spell check was a novelty.

Edit: Just noticed. 70’s? Really? That’s very impressive. I’d be very interested to see an example.




That’s very interesting, thanks!

I would say this is perfectly normal. The IME can’t always guess the correct word if you’re typing a full complex conjugation form (causative-passive of intransitive verbs comes to mind) and sometimes there will be different valid kanji combinations for the same phonetics.

Which IME are you using specifically? There is a couple of them, I think.

We actually had a discussion about this in a different thread and someone mentioned that Japanese people also use regular QWERTY keyboards with IMEs, because they’re easier to handle, but the way they type might be a little different. Still the result is the same.

The few times I saw anime characters typing on PCs it also seemed like they’re relying heavily on IMEs.

Except in this case the system doesn’t even need to use machine learning or anything elaborate. A simple scoring algorithm will prioritize kanji + kana combinations which you use more often :slightly_smiling_face: .

I actually realized today that once you get into the habit of writing shorter phrases and cycling through suggested kanji combinations, it works the same way as typing with T9 on your phone.


IMEs usually respond better when you convert more frequently. It gives you more control over which kanji is used for each semantic block (one vocab word plus prefix/suffix/particle if applicable), since you can cycle through different options for each one. Also, sometimes pressing tab instead of space will give you different choices. On windows you have the choice between microsoft IME (built in) or google (you have to go download it.) I prefer the google one because the Microsoft one often hangs or is unresponsive on my computer but maybe that’s just me.


Like this:

This is the ALPS CP10SJ550A Kanji Keyboard, which was in use before kana-to-kanji conversion was invented. Having trouble determining the exact timeframe for this, though.


At least for me on my Macbook, it offers a list of possible kanji you might mean.

I have heard that even japanese typers will often do this to force certain kanji, or type a kunyomi reading instead of an onyomi or vice versa to get a kanji, even if the word isn’t going to be pronounced with that sound.


I read a super fascinating book called The Chinese Typewriter about solutions for typing Chinese characters in the typewriter era. I’d definitely recommend it if at all curious.

The gist as I remember, is there were solutions invented and used successfully that used banks of key groupings as a kind of analog predictive text.


I’ve really only found that IMEs can’t guess what kanji I want when it’s a particularly obscure one.

For most common kanji, they should be there somewhere even if you have to scroll a lot.

Most of the time there’s not much reason to type isolated kanji outside of words. Unless, I suppose, you’re someone like us, frequently discussing kanji studies.

For Japanese people, I suppose the most common situation for getting isolated kanji is when typing names with unusual readings.

1 Like

Discovered one example of this recently - うたる is a Kansai-ben version of 買ってあげる. When trying to Google for what it meant, I came across one forum where a user was asking “people of Kansai, how do you type this?” - the response was “type かう and convert, type たる and don’t convert.”


That keyboard is scary!!! O_O