HuaRenCa
Back to Forum
Community

WeChat Keyboard Sensitive Word Extractor

醉听松风
醉听松风

2 weeks ago

Saw this post — turns out it's not that I'm clumsy, it's that Cook is playing tug-of-war with me in the background every day! — #13, from xjx, inspired me,

Actually, WeChat Keyboard also has sensitive words, and they're absolutely hilarious

Spoiler

For those without a v2ex account, I've posted them for you

Out of interest, with AI's help, I wrote this WeChat Keyboard Sensitive Word Extractor

The word list currently has just over 10,000 entries. Preview of the extraction results:

(I won't post the results here since the content is too explosive)

This tool extracts from the .\BuildIn\cell_dict.ssrc.zh file in the WeChat Keyboard installation directory. Tested and working on the latest Windows version

Usage:

python export_sensitive_single.py -f <cell_dict.ssrc.zh> -o <out.txt>

There are also some interesting findings: WeChat Keyboard has another built-in, less sensitive "sensitive word" list (kind=440)

Although this extra word list can be used for word composition, it's categorized separately in the keyboard implementation.

As for whether there's any special handling (such as logging), I haven't analyzed further

Tool download link: https://pastebin.com/emEXbrCh (extraction code: x7E56zQ4uZ)

For those who don't want to download, the full word list is provided

WeChat Keyboard Sensitive Word Extractor

6
1.7K

Comments (6)

Your avatar
Sign in to comment
荒野丛林
荒野丛林2 weeks ago

The path is probably different on macOS, so it's safer to confirm which system you're on before making any changes.

易济
易济2 weeks ago

I'd like to ask roughly how long it will take to extract over ten thousand entries, and whether I need to completely exit the input method first. Also, for the kind=440 part, are you planning to export that as well, or will you only process the main dictionary?

飘移不过弯
飘移不过弯2 weeks ago

Regarding your point about kind=440 being categorized separately—I'm wondering if it's just used by the input method for phrase association and has nothing to do with the interception logic. Could we confirm this from the file structure?

adas67890
adas678901 week ago

Verify the hash first—don't just double-click and run the script.

猫科斯塔灬
猫科斯塔灬1 week ago

Be careful when using tools like this on a shared computer—exported txt files can be picked up by sync drives or backup software. Also, the install directory changes between versions, so file paths may break after an upgrade. It's best to run it in a virtual machine or isolated environment, and don't casually share the results with others.

平安江南咏
平安江南咏1 week ago

The core issue is that if the dictionary source and version don't match, the results will be distorted. It's recommended to note the input method version number used during extraction, so it's easier to compare and update later.