Learning to Type Chinese using Cangjie - a write up

[Tomsima]

Disclaimer: This write up is not a guide on how to type using Cangjie, check out the wiki page for a basic intro if you're interested. This is aimed at anyone who simply wants to know whether learning a new input method is or is not worth the time investment.

 

2020 has been a very strange year for me, as I'm sure it has for most of us. With all the extra time, I decided to get down to some things that I've wanted to do for a while but...just never had the time. One of those things was learning to type Cangjie both fast enough that I can use it for live conversation on Wechat, and for practicing my character retention abilities.

 

There are a number of shape-based input methods for Chinese out there, the most famous being Cangjie (倉頡), Dayi (大易) and Zhengma (鄭碼) for traditional, and Wubi (五笔) for simplified. I chose to learn Cangjie as it is well suited for typing both traditional and simplified, which can't be said of most other shape-based methods (most are now able to some extent, but mainly rely on 'conversion' rather than directly typing in the specific character according to its structure). Thats not to say Cangjie is 'the best' of these systems, its just the one that suited my needs the most. Other benefits of Cangjie are that it is widely available and license-free, so no worries that it will suddenly disappear or require some payment to use. It also uses a lot less keys than methods such as Dayi, so less finger stretching. Regardless, I believe Cangjie is an incredibly well-designed system, a real work of genius that functions to break down computer-font characters in the same way stroke order helps with handwriting characters.

 

After 6 months of practice I have racked up just close to 100 hours of typing practice on anki (typing out sentences from memory based on prompts). I can now reach around 25-30cpm. I type at around 60-70wpm in English, so I've still got a long way to go, but I'm happy with my progress as it stands.

 

Here's what I've found is important on my journey:

 

1. Your keyboard keys affect how a shape-based input method helps with character retention

 

I originally set out using normal keys with alphanumeric symbols.

 

I learned to touch type fairly quickly in Cangjie, but found that I began to see characters as strings of English letters in my head, a little like how when you're typing in pinyin you often think of the romanised version of what you're writing before the image of the character floats into your mind. This became quite annoying and counterconstructive, so I got some Cangjie stickers from ebay and stuck them on blank keycaps to see what difference there might be.

 

 

The difference was noticeable immediately, as I began to associate the keys with Chinese characters much quicker. However, I still found that with some of the more difficult keys (where the character and the element it could represent are connected in a fairly abstract way), my brain would start remembering the string of keys for the character instead of properly decomposing it into its elements. The brain always chooses the easiest option I guess. A good example of this would be 麼, where 戈 represents both 广 and 丶 in the decomposition,  with 女 also representing the stroke ?, it was just easier to remember 麼=戈木女戈, or even just the shape the keys made on the keyboard.

 

So I decided to make a set of keys similar to the ones you see for 五笔, where every single symbol is listed on the keycaps (ive seen them for 鄭碼 too, probably because the amount you need to remember for it is too much of a burden on the brain).

 

 

I should emphasise, I decided to use this keyboard specifically for the purposes of character retention. If I wanted raw speed I would just use blank keycaps and rely on muscle memory. This keyboard has had a massive effect on how Cangjie has helped with remembering character writing, and if anyone is interested I'll be happy to send on the inkscape file. Now when I look at my keyboard to type 麼 I can actually look for 广 - 木 (-木) -?-厶 instead of remembering some arbitrary code or pattern.

 

 

Think that looks scary? Its not, it is very intuitive and can be learnt in half an hour of typing I would estimate. Check out 徐碼 for a typing system that has a single code for every single character you could possibly type. Bet you like the look of that Cangjie keyboard now:

 

 

2. Cangjie 5 is a massive improvement on Cangjie 3. Microsoft Cangjie is riddled with errors.

 

I first set out using Cangjie probably around 2 years ago, but it was only really out of curiousity and I only used it on my phone. I didnt realise it at the time but I was using the 3rd generation of the system (for reference, 1 and 2 were largely just glorified betas). Then when I moved onto using cangjie on my laptop (ms surface), I discovered that many of the codes were different, despite it still being classed as Cangjie 3.

 

Thankfully I came across this fantastic wikibook which not only explained the errors that MS has made in its own hacky version of Cangjie (after parting ways with the creator of Cangjie), but also showed how the 5th generation of Cangjie had corrected all the weird decomposition errors and inconsistencies in Cangjie 3. I immediately switched to Cangjie 5 and have not looked back, it is internally consistent and logical throughout. I strongly recommend any future students of Cangjie to use Cangjie 5, it is a pleasure to type with and really feels like you're writing characters, just like that feeling you get when you type English and your thoughts seem to just 'appear' on the screen - there is no feeling of detachment.

 

Here are some notes I made when I first made the switch from MS Cangjie 3 to Cangjie 5 (using 倉頡平台)

 

• Correction of character selection order based on frequency. Eg 致 before 玫，知 before 佑. Damn that ms input was annoying, always having to add in '2' after so many common characters.
• recognition of 尸 as representative of the double dot, eg 假 人口尸水
• 應：戈人土心 this is fantastic, finally the parts are separated properly!
• 篼 has been corrected to 竹竹女山 (instead of 竹竹尸弓, which breaks away from the treatment of 兜 as a single unit (both in 3 and 5)
• 撐 and 撑 have their own unique codes (another MS error, typing 牙 here gives you 手...)
• 木廿 来
• 大木 东
• etc the list goes on...

 

I encourage anyone thats interested in comparing the differences between CJ 3 and 5 to have a look at this list. In fact, browse through the whole book, its incredibly well written. (Written by the 'boss'? of 倉頡之友, a forum without which I would never have found any success in learning 倉頡).

 

3. Cangjie is really fun to type with

 

If you've ever felt the frustration of having to cycle through pages of characters to find the one you want, hate typing out whole words then delete the parts you don't want, or if you just can't stand 联系and 练习 causing all your friends to question what on earth you've been doing with all those hours of Chinese study, then Cangjie is defintely worth a try (or any other shape-based input method for that matter).

 

Once you get used to typing using a shape-based method, you realise just how annoying typing phonetically is. Yes, I get it, its very, very, very easy to learn, and it means you don't have to remember how to write characters, only recognise them. But if you are at all interested in writing Chinese, then try Cangjie (or Wubi if you're simplified only gang) and I'm sure you'll never look back. There is nothing more satisfying than seeing an obscure character and being able to check it instantaneously in your dictionary. I still remember the first time I saw 鑾  and realised it was just three keys right next to each other (女火金), the pure satisfaction...

 

Here is an update video of me typing from today：https://youtu.be/DaZ9QRSKTbc

 

I drafted a short paragraph then recorded myself typing it back out. There are errors, and its pretty slow going, but still, shows where I am honestly at after 6 months.

 

Hope some of this helps, and if you've got any questions let me know and I'll try and help out

[kingwithin]

how does it compare to writing practice regarding character/pronunciation association? I find that one of my major drawbacks with typing mandarin is having to spell it out in romanized characters and always associating the sound itself with the pinyin rather than the character

[Tomsima]

are you a heritage learner? what sort of stage in your learning are you at? I can only speak from a native English speakers point of view, but learning a pronounciation based input method came very naturally to me, whereas cangjie I have had to consciously learn it with sustained effort (although this would be true of a native speaker to some extent)

[kingwithin]

Australian, native language is english, I'm currently self studying mandarin at a fairly low level, hoping to improve considerably by the time borders open up again

[Tomsima]

In that case I would presume you'll see fairly rapid progress as you get more familiar with pinyin. But if you feel more confident writing with characters from an early stage thats also great!

[Yadang]

On 6/27/2020 at 11:58 AM, Tomsima said:

I should emphasise, I decided to use this keyboard specifically for the purposes of character retention. If I wanted raw speed I would just use blank keycaps and rely on muscle memory. This keyboard has had a massive effect on how Cangjie has helped with remembering character writing

 

I'm curious whether you think it would be beneficial for one to use Cangjie if their end goal is reading but not writing. Would it help with character recognition? Is it possible to use Cangjie without fully knowing how to write the character you're trying to type? 

Really, what I'm getting at, is that I don't care about writing (by hand), but do care about reading, but have some problems with character recognition for some characters. I'm wondering if typing in Cangjie would force me to be a bit more mindful of the characters (clearing up some of the problems with character recognition), without requiring me to learn to write all of the characters (which would be a more extreme way to clear up my problems with some characters). Any thoughts?

[imron]

19 hours ago, kingwithin said:

Australian, native language is english, I'm currently self studying mandarin at a fairly low level,

My experience with Wubi (similar to CangJie in that it's based on shapes rather than sounds) is that using a shape based method will not be very effective unless you already know over a thousand characters, and preferably over 2,000.  If you don't know enough characters it will be much more difficult to build up proficiency in the input method.

 

8 hours ago, Yadang said:

I'm wondering if typing in Cangjie would force me to be a bit more mindful of the characters

Once again speaking of my experience with Wubi, I believe it will.

[Tomsima]

On 7/29/2020 at 4:44 PM, Yadang said:

I'm curious whether you think it would be beneficial for one to use Cangjie if their end goal is reading but not writing.

 

I mainly use cangjie to type out new words and phrases from my daily reading. When I come across a new word or interesting turn of phrase, I will make an anki card with an input field for Chinese. I find this really helps me to distinguish lesser seen characters far better.

 

I agree with Imron. I had a base of probably 4000 or so characters I could already read and write by hand before starting, so it would be fairly fruitless for me to speculate on how useful shape-based typing systems are for learning characters without knowing how to write by hand. However, at the very least, I can confirm that using cangjie makes my brain pay greater attention to the contents of each character  

[5537257165Telegram]

I cannot agree more to the point of improving characters retention.My goal exactly. I am still struggling with many characters. Like why a frequent character as  

[calibre2001]

There's a comment on MS Cangjie 3 vs 5 where the latter is reported to be a significant improvement.

 

To supplement, the majority of Cangjie users (using Windows presumably) are Generation 3 users since they grew up using this unknowingly. Many aren't even aware of 3 vs 5. The other issue is in Windows 10, the default Cangjie is generation 3 only. Microsoft hasn't taken steps to rectify this by including a generation 5 option despite user requests. Thus its still probably safer to learn MS Cangjie 3 for practical reasons.

[flashytiger]

Quote

Disclaimer: This write up is not a guide on how to type using Cangjie, check out the wiki page for a basic intro if you're interested.

Could you link to the wiki page you're referring to? I'd like to know what resources you used to start out.

[Tomsima]

@flashytiger quite literally the Wikipedia page for cangjie 

https://en.m.wikipedia.org/wiki/Cangjie_input_method

[Tomsima]

Thought this might be useful for any beginner Cangjie learners. I usually type on my own computer, but recently have had to use a work computer a lot, which means im stuck with the annoying, error-ridden 'MS Changjie' IME. I made a list of these as a cheat sheet for myself for typing in simplified using Cangjie. As Cangjie was originally designed for traditional and only later adapted for simplified, discrepancies have arisen between CJ3, 5 and MS Changjie. Here it is typed up:

 

Component decomposition differences which arise from writing simplified components in Cangjie using MS Changjie (as opposed to 5) are:

• 冫   = 卜 (not 戈一. This is used even in characters like 习 (尸卜 rather than 尸戈一)
• 丬   =  卜中 (not 中一. For the same reason as 冫)
• ?   =  戈月 (not 火, eg. 学: 戈月弓木 not 火月弓木. Note that 龸 is still treated as 火月 in MS Changjie)
• 讠   =  戈弓 (not 戈女)
• 产   =  卜廿 (not 卜竹)
• 业   =  廿一 (not 廿金)
• リ   =  戈竹 (not 中中) eg. 师 is 戈竹一中月 (not 中中一月)

 

Wider errors are:

• Any character with 钅will randomly only accept 金 rather than 人心 as its code (eg. 钱  =  金戈一 not 人心戈十)
• Any character with 发 on the right hand side is given the code 女戈水 (not 戈女水) despite the fact that MS Changjie actually uses 戈女大水 for 发 itself

 

There are other errors, but these are mostly for very rare characters. There are some common characters which have two codes assigned, one the original error, and one the later correction. As you might have spotted, the key difference in the systems is how MS Changjie treats dots (ie it treats them wrongly...)

 

[etatoby]

On 6/27/2020 at 8:58 PM, Tomsima said:

I immediately switched to Cangjie 5 and have not looked back, it is internally consistent and logical throughout. I strongly recommend any future students of Cangjie to use Cangjie 5, it is a pleasure to type with and really feels like you're writing characters

 

Can anybody please tell me where can I download Cangjie 5 for Windows 10? I looked around but I can't seem to find it. On Android it's a simple option in Gboard.

[Tomsima]

Link is in the original post, 倉頡之友 has since released an updated 2022 version 《倉頡平台2022》which works on both Windows 10 and 11. I have used 倉頡平台 for many years now for typing Cangjie 5, can recommend.