14 comments

[ 1.4 ms ] story [ 40.7 ms ] thread
That's really cool! I got very frustrated with SwiftKey when I tried to use it to text my Chinese friends.
(comment deleted)
not that this would actually help anyone in need of a transliterator, but probably of interest: mit press is publishing a book later this fall on how Chinese characters triumphed over the QWERTY keyboard and laid the foundation for China’s information technology successes. full disclosure: i used to work at the press. got to read some of the manuscript, and it was indeed fascinating: https://mitpress.mit.edu/books/chinese-typewriter
While this is a cool research project to learn Tensorflow, it's no where near the competition. All Apple iOS / Microsoft Windows / Google Android's stock pinyin input methods has more or less an AI component to automatically predict the Chinese characters on the fly while you type - they work extremely well. There are a about close to a hundred other Chinese input method vendors in the market right now other than the top 3 tech companies' stock Chinese input methods. Leading the bunch is a company called iFlytek[1], which utilize voice recognition AI to improve input speed.

And yes I am a native speaker and use pinyin on a daily basis.

[1] https://en.wikipedia.org/wiki/IFlytek

Aren't these non phonetic writing systems a great time loss in general? Still wondering if Japan/China/similar would not get a great economical boost from ditching their non decipherable ideograms. Just thinking of the time those poor students spend remembering them instead of learning useful math/history/foreign_language just pains me.
The characters are a reflection of these societies' history and culture. Doing away with them would break the connection future generations have with their own literature, art, and history.

China did officially approve a new system of simplified characters (簡體字) in the 1950s and 60s to improve literacy levels and save time (1), but many people think the thousands of invented characters don't look nearly as good as their traditional counterparts (繁體字) which are still used in Taiwan and Hong Kong and some overseas Chinese communities in SE Asia.

There is also a slight learning curve for people from China to learn traditional characters and Taiwanese and Hong Kongers to learn the simplified versions used in China, but the exchange of mass media, social media, and Chinese travellers seems to have helped people in the different areas get more familiar with the different systems.

1. https://harvardextended.blogspot.com/2007/11/undoing-simplif...

Why would these people change to a phonetic writing system when most of the world's isn't? English for example is a morpho-phonemic one. The Chinese writing is also more resilient to language change, and given proper education, texts from millenia can still be read.

If you look at the PISA scores[1] for example, the top 10 includes Singapore, Japan, Taiwan, Macao and Honk-Kong. So the Chinese script and its derivatives is clearly not impairing its users to learn maths.

On the foreign language side, it's even a boon because someone knowing a CJK language wanting to learn another one can rely on a lot of words with shared roots to learn it faster. But as the phonology of these language is different, the knowledge of the Chinese characters can help connects the dots between words, which aren't obviously related when romanized.

[1] https://www.oecd.org/pisa/pisa-2015-results-in-focus.pdf

> Why would these people change to a phonetic writing system when most of the world's isn't? English for example is a morpho-phonemic one.

This isn't some binary category where you have or haven't achieved holiness. Compared to Chinese, English is indistinguishable from a purely phonetic script.

This is still wrong. First, because English taken for itself is nowhere near a phonetic script. It uses a lot of graphemes to represent the phonemes of the language and there is no bijection between the two sets (as expected for a phonetic system). In the end, the number of combinations to memorize for reading perfectly is probably in the hundred or multiple hundreds.

French has the same problem. It takes years to train natives to write to perfection; and the non-phonetic nature of the script bites hard the foreign learners.

Second, at its core the Chinese writing evolved from a limited set of symbols to represent syllabes. The current system is still mainly based on it and the knowledge of around 400 components can help read and memorize of lot more characters. The phonetic nature of the Chinese script is not apparent to people who doesn't studied it, but it exists.

The phonetic nature of the Chinese script is so weak that in the foreword to https://www.amazon.com/Xunzi-Complete-Text/dp/0691169314/ , the author observes that while he uses his own translations, he is forced to defer to existing scholarship as to what parts of the text rhyme, not being qualified to determine that for himself.
The Chinese are very proud of their writing system - to the point that Chinese calligraphy is pretty much high art there. We love it when non-Chinese talk about how hard it is to learn - it gives that quick ego-boost like what you get when someone talks about how hard it is to learn something you already know.

Sometimes I think part of the reason China/Japan is so inward-looking is because they've invested a LOT of time into learning the ways of their own culture.

And British is trying to use Chinese textbooks to boost math education. Non-phonetic doesn’t mean easier to grasp the words meanings. Try read a Taxonomy or anatomy book and tell us whether you need sometime to learn each words.
This is very interesting.

I tried to use swiftkey chinese beta many years ago, but couldn't accept it because it only allowed pinyin input with simplified characters.

Currently I use Google Zhuyin keyboard (which also has a pinyin mode) which does pretty accurate prediction and has a unified interface with my cantonese (yale) and korean keyboards.