Chinese input method for teenager

This one is not bad either.

1 Like

What input method do people use in workplaces here? Don’t tell me they blunder about with zhuyin? No wonder email is not a thing in Taiwan…

There is an IME out there called 追音, can’t find its English name, but if I were to name it, I guess I’d call it Catchup Phonetics. It is a popular IME among court reporters because with practice one can easily go above 150 characters per minute.

The product is a result of the Judicial Yuan asking the makers of Natural Zhuyin (the same people that came up with the Hsu keyboard layout) to make them a workable stenotype like IME.

The keyboard layout looks like this:

The big difference is, instead of pressing one button at a time, you are required to press at most 3 buttons at once. So each syllable is entered with one motion, and AI will help picking out the correct characters in a sentence.

It’s fast, but also sounds carpal tunnel inducing.

I found an example to demonstrate how carpal tunnel inducing it can be. If you want to type 不會, you need to first press QJF at the same time, and then CH, at the same time. Give that a try, and you can imagine what that would do to your hand if you quickly switch between these awkward gestures.

1 Like

(just an aside from the OP, I truly appreaciate all the replies here. It’s going to take me a few weeks to get back to this thread, but I will do that)

Or maybe a few months :slight_smile:

Looking at the Boshiamy website (with my bad chinese), this is a paid service, right?

But there doesn’t appear to be a download (or instructions) for Linux (?).
Google suggests that folk are installing using fcitx (sudo apt install fcitx-table-boshiamy), but is this actually Boshiamy, or just something similar that ‘kind of’ works on Linux?

Am confused and advice would be appreciated.

RE: paid service: sort of. Chinese input methods relying on graphical decomposition (e.g. Boshiamy, Cangjie, Wubi, Zhengma, etc.) are all at the end of the day just a mapping behind a sequence of keys to a (maybe unique, depends on the input method) Chinese character.

So, given a common input engine (e.g. fcitx), a simple text file that acts as a dictionary listing all of these mappings (e.g. one line might be “ab 明”) is sufficient to yield the input method.

So in this sense the Debian package you’re looking at is the “real Boshiamy.” Now Boshiamy hasn’t traditionally seen that much usage in the Linux world because, as you’ve found, it is (or was) protected under patent. Its patent should’ve run out about a decade ago and that’s why it’s been included in Linux.

The predominant Chinese graphical decomposition input method in the Linux world is Cangjie, whose creator effectively released it into the public domain in the early 1980s.

If English is your main working language, the unfortunate part about a lot of these graphical input methods is that their learning resources tend to be in Chinese; their English equivalents are generally very scattered and fragmentary.

N.B. I have a bias towards and invested interested in Cangjie, because I happen to be working on a website that aims to be a comprehensive English resource and typing practice for Cangjie. However, I don’t know what this forum’s policies on promotion of links are; if a mod is looking, I’d appreciate some guidance here.


We are A-OK with promotion of helpful things. Not so much with promotions of yet-another-random-Mainlander’s-learn-Chinese-with-me site, especially if it’s behind a paywall. Yours isn’t that. :slight_smile:

If you ever decide to make a big business out of it (good luck, with Cangjie! not trying to jinx you but still…) you might want to think about purchasing an ad here, but I would assume that’s still quite a ways off. Meanwhile, your link is helping the community and we encourage that! 加油!

1 Like

I use Boshiamy exclusively on Linux. My goto IME framework is gcin.

1 Like

Thanks for the replies.

fargo, please fell free to post that link, then. Suspect this could be a steep learning curve, but I’ll continue when I find the time.

I feel that any input system that requires a keyboard will definitely force you to enter the correct tone. This, in my opinion, is a good thing. I’ve gotten so used to predictive texting on smart phones that I don’t even really pay attention to tones and just let the predictive texting take control and I will select the correct character based on the one or two simple phonetic symbols of that I enter.

Using predicting text (in any language) doesn’t help very much with my writing abilities, so any method that requires use of tones in typing should be advantageous for every aspect of language learning. Heck, my English spelling has also suffered as a result of auto-correct.

Alright, then; I’ve made a separate post here that goes into my thoughts in a bit more detail:

But the link itself is: At the moment I’m still trying to canvas interest to figure out how much time I should devote to this project (basically the more people interested the more time I’ll spend).

Ha; oh I doubt it’ll ever be a big business. But it may yet turn out to be a fun project nonetheless!

Could you briefly outline how cangjie differs from boshiamy? They both seem to operate on the same principle - character decomposition, then assigning a letter to each of the components.

To insteall gcin either simply run

sudo apt install gcin

Or get the newest version by adding the maintainer’s repo:

sudo apt-key adv --keyserver --recv-keys 835AB0E3
echo “deb eliu release” | sudo tee -a /etc/apt/sources.list
sudo apt update
sudo apt install gcin

You will have to open Ubuntu or whatever Linux distro’s setting menu, navigate to the language tab and select gcin as your default IME

You can of course use the default ibus by building your own Boshiamy database file, but I have never gotten used to ibus. It’s slow and written plus maintained by some Chinese dude. Gcin is written and maintained by Taiwanese.

Warning: longer post than I intended. Also standard disclaimer: since I’m working on an online resource to teach Cangjie, I’ll naturally be biased towards Cangjie, but I’ll try to keep this even-handed.

Sure! You’re exactly right; at the end of the day they each have a set of “radicals” (to differentiate them from the usual Chinese radicals I prefer to call them fragments) that they then map onto the usual 26 keys of a keyboard. They then use these to piece together a character. Because both systems use a set of fragments that exceeds 26 in number, both are forced to group a set of fragments to each key.

Hence there’s differences in four places between Boshiamy and Cangjie:

  1. The base set of fragments
  2. How they decide to group fragments by keys
  3. Choice of determinism
  4. Ecosystem

At a high level the trade-offs between the two systems all boil down to different decisions made in each of these dimensions.

The base set of fragments

Boshiamy tries to keep its base set of fragments as similar to existng Chinese characters as possible; it sometimes compromises and uses incomplete fragments, but many are just plain Chinese characters. Cangjie doesn’t and essentially re-invents its base set from the ground-up. On the one hand, this makes Boshiamy’s base set of fragments much less intimidating to a new user than Cangjie’s which truly look like alien, incomplete fragments of characters. For example, compare Boshiamy’s table of base fragments with Cangjie’s

On the other hand, this means Boshiamy sometimes has to “stretch” some of its base set of fragments to cover certain cases. For example, the top of 置 is a 四 (“okay yeah if I straighten and extend out the lines in the middle so they touch the bottom it’s a 四”). Cangjie’s base set of fragments, while alien, is chosen so that it has to do essentially no “stretching” of this sort. Cangjie’s choice to re-invent its base set also allows it to get away with fewer base fragments (101) than Boshiamy (341).

Moreover, Boshiamy’s choice to try to keep its base radicals similar to pre-existing Chinese characters means that it has different “modes” for simplified and traditional characters, which you can enter on a character-by-character basis to write mixed traditional-simplified documents.

Cangjie’s choice of base fragments is meant to encompass all known character variants. And so there is no separate simplified or traditional mode. It’s all just the same rules for breaking down a character. Because simplified characters look different than traditional characters this means a simplified character and traditional character for the same word will have different key sequences.

Grouping decision

Boshiamy uses a combined system of Chinese pronunciation, English translated pronunciation, Chinese morphological structure, and rote memorization to group its base fragments to a key. For example the Boshiamy base fragment 刀 has a “d” sound so it gets mapped to the “d” key. Note that this pronunciation system does not correspond to Pinyin (I don’t think it corresponds to any standardized Romanization at all). For example 升 maps to “x” also by its first sound. 也 maps to “a” because its entire word sounds like “a” read out loud. 王 is mapped to “k” because 王 means “king” in English and “king” begins with “k” (whereas e.g. 开 is also mapped to “k,” but based on its Chinese pronunciation). 佳 maps to “v” because it translates to “very good.” As for Chinese morphological structure 工 looks like “I” so it’s mapped to “i.” And some are just rote memorization (欠 is also mapped to “i” just because).

Cangjie groups its fragments entirely using morphological structure and then essentially arbitrarily assigns each group to a key (you’ll often see Cangjie keyboards relabeled with each key having a single representative element from that group).

Choice of determinism

Boshiamy significantly reduces the chance of collision compared to Pinyin or Zhuyin, but every so often still requires a user to slow down and disambiguate between different characters that map to the same sequence of keys.

Cangjie on the other hand is fully deterministic. Every Cangjie character has a unique sequence of keys assigned to it where structurally similar characters are distinguished using a disambiguation key (usually “x”). E.g. 暈 and 暉 can both be decomposed as “abjj” according to Cangjie decomposition principles. And if you type in “abjj” usually your IME will give you a drop down menu where the default option is 暈 (i.e. that is the character that will be entered if you continue typing), but where 暉 is also displayed. However, that list is ordered by relative frequency of that character in Chinese and “xabjj” will unambiguously give you 暉 (and is the “canonical” representation of that character in Cangjie so to speak) so that assuming you’ve memorized this, you can type without ever having to resort to a drop-down menu.

Ecosystem support

Neither Boshiamy nor Cangjie see much use in mainland China (the predominant graphical input method there is Wubi). Cangjie is popular in Hong Kong (along with an abbreviated variant of it known as Quick Cangjie that trades off determinism for fewer keystrokes). Hong Kong is probably the only place where graphical decomposition systems are more popular than phonetic systems (Pinyin is hard since Mandarin is rarely a Hong Konger’s first language and Jyutping has had only a minority of people adopt as far as I can tell). I’ve never seen a true survey done, but I think Boshiamy and Cangjie are essentially equally popular in Taiwan (but far outpaced by the popularity of Zhuyin).

Finally, from an end user’s point of view Boshiamy’s history as a proprietary product rather than Cangjie’s history as one in the public domain has tradeoffs for both sides. On Cangjie’s side you have more systems that support Cangjie because they weren’t afraid of getting sued (although Boshiamy is supported on enough systems at this point, especially given that its patent has expired, that this is much less of an issue these days). However, in Boshiamy’s favor, because Cangjie is public domain and therefore free to tweak and its creator is quite meticulous and issues updates to the system as he thinks of improvements, small variants of it have appeared over the years (Cangjie 3, Cangjie 5, etc.). This can make it a bit confusing for a newcomer to figure out which variant of Cangjie he/she should use and can make some of the learning materials slightly wrong for the variant you’re using (another reason why I wanted to make a single comprehensive Cangjie resource). For Boshiamy in contrast, on Mac, iPhone, Android, and PC there is an identical Boshiamy system that is produced by the single company behind Boshiamy.

Parting Words

That was a bit long… At the end of the day I suspect the learning curve between the two despite these differences is probably quite similar. I say suspect because I’ve never become fluent in Boshiamy, but I see no reason it would be substantially harder or easier (the boost from its mnemonics and pain from its ad hoc-ness probably come out as a wash) than Cangjie. Given the right materials, it’ll be on the order of weeks of learning to learn either.

When I was making my own personal journey to select a graphical decomposition input, I decided on Cangjie mainly due to its full determinism and the generous choice of its creator to put it in the public domain. I figured that since I was going to put the time in to learn a graphical input system to reduce the amount of times I have to deal with a popup window and choose some selection, I might as well go all the way and learn a system where I never have to deal with a popup window if I so choose. Make of that what you will.

1 Like

it was a bit long. But I’ll get round to reading it :-).

Boshiamy has a 4 key maximum per character, unless you run into a really rare character that requires selection after 4 keys, you would never have to type more than 4 Keys per character.

Boshiamy maps a bunch of commonly used characters to just 1 or 2 key combinations, so most of the time you’d never even have to type more than 3 keys. Whenever I used more than 4 keys, or need to select a charater, I start to wonder if there is a simpler way to type.

If you work with classical texts, you’s know with UNICODE CJK extension sets, there are so many characters, there is no way to have a unique key combination to represent all of them. Character selection is inevidable. The important thing is that all the commonly used characters have a unique key combination, in which they are the default output.

Cangjie on the other hand has a maximum of 5 keys per character, with most characters up in that 5 keys range.

Also, even though Cangjie stays within the 26 alphabet keys on the keyboard, you have to remember a whole new sets of representative fragments (one character that represents a bunch of nameless fragments) and where they are on the keyboard. Most of the time Cangjie typers remember how each character is typed by the sequence of these representative fragments, instead of the key on the keyboard. That’s why new Cangjie typers would need to have those keys marked on the keyboard.

Boshiamy uses the alphabetto represent fragments, and most fragments are actual Chinese characters. In that way it’s much easier to type, and the muscle memory is easily transferable between English and Boshiamy.

As for the memorization part, I think they are equally difficult. At the end of the day, you just have to memorize them and commit it to muscle memory to type efficiently.

Cangjie’s biggest advantage is that it comes with every operating system that supports Chinese. Although as mentioned, there are version differences that might trip you up. Boshiamy is available on all platforms, and you don’t have to get the software from the LIU company. You just have to know how and where to get it.


Had a friend who wrote a book in English on the Cangjie system- he published it just as the Zhuyin tide swept in.

I’m native and I use 新注音 on windows or 注音 on mac. Some say 倉頡 or 無蝦米 works faster but unless your work requires a lot of keying in I don’t think it’s necessary to learn a whole new system. Plus, school computers suck and might not have input system built in other than 新注音.

If you feel like spelling system for mandarin (pinyin or zhuin) is crap that’s probably because mandarin wasn’t designed that way. Spelling system for mandarin didn’t exist until 1918 and it had a political or even racial reason.

I think you’ll find “spelling systems for Mandarin” predate 1918 by quite a few years.