We need a translator for "hate speech" to "obfuscated truth". I got in trouble on YouTube every week or so (24 hour bans) for using certain words. Instead of those words I now write "female canine", "bovine excrement" etc.
I want an Android keyboard that automatically replaces such terms with their obfuscated counterparts. It could also work with simple character replacement. B1tch. Bu11sh1t. Etc.
Well, the first option is really good, the "bovine excrement" is hard to pin down without being draconian.
Any variation of a word and it's l33t5p3@k counterparts can be easily detected with regular expressions or regex as we call them in the computing world.
This is some example code that shows how this can happen, using python:
import re
This imports the 'regular expression library in python; basically makes it available to us for use as pre-written code in the current code we want to run.
word_list = ['farts', 'poop', 'butts']
Let's use a list of words. This is how a list is created in python [ ] mean list, and each of the entries needs to be quoted, single or double, and comma separated, this is so the compiler (what takes code like this and makes it machine code) can understand it.
This is the secret sauce here. Basically we have other FASTER methods to go through a set of characters, but for this simple example, we have defined a key:value pair. Kind of like firstname: John, lastname: Doe. This construct leetspeak_dict is a dictionary, and they just contain the k:v pairs. So when 'a' is encountered, later on we will check 'a' against a capital A, @, and 4, which are all common substitutions.
# Create a regex pattern for each word in the list
patterns = [re.compile('|'.join('(?:{})'.format(leetspeak_dict.get(char, char)) for char in word)) for word in word_list]
This part just pre-compiles the patterns for using a for loop for each character in the words from the word_list that we defined earlier.
You could test this using re.match, perhaps even re.fullmatch, but they are fiddly at times. This is just to illustrate.
TL;DR: Your first approach makes pinning down your speech much, much harder. Why? I have to filter so many words that regular speech breaks down, and it it gets exponentially more expensive to police your speech.
Character substitution is easy to detect and block.Huge word lists introduce overhead in checking for disallowed words, and it's even worse with phrases in doing a version of the above, typically in something like javascript. Python is a lot easier to understand, so I did python here. You make them pay a lot more money to censor phrases too. So much so that eventually it becomes too much. If everyone does this, the only choice is manual content moderation, which in the event of bans, drives revenue off the platform, and costs them actual money. Even AI cannot reliably do this yet, because we change the euphemisms constantly. The Euphemistic Treadmill, that liberals un-ironically use to police speech (aka Retard/Austist -> Neuro-Divergent Person) becomes the reason that they, and AI can't control it. Conservative views are worth money too.
Not only that. You now make normal conversations subject to scrutiny, which hurts the credibility of the platform with normie users. False positives will pee off normies.
Regular expression matching is exponentially harder and more expensive with additional phrases, and creates false positives that will be impossible to police without affecting normies.
WEW LAD! How many cups of coffee did you drink this mornin fren?
While I appreciate your passion, a quick 1 minute search would have led you to this document that shows the hype in OP's video is much ado about nothing.
Wow and boy did you call it! I had like 3 cups in one GIANT thermos this morning!
I saw it was about section 5 last night, that targets campaigns and all, but I see the writing on the wall, after this it's a small text update to remove campaign and were all on a terror watchlist for calling something obviously gay fruity.
I'm paying them 297.00 per month atm. So I am just ready to get down to 40 per month with an MVNO (I work in the industry and know better, but have had to have the Tmo home internet in the area for my brother, so paying out the nose)
We need a translator for "hate speech" to "obfuscated truth". I got in trouble on YouTube every week or so (24 hour bans) for using certain words. Instead of those words I now write "female canine", "bovine excrement" etc.
I want an Android keyboard that automatically replaces such terms with their obfuscated counterparts. It could also work with simple character replacement. B1tch. Bu11sh1t. Etc.
Whore doesnt work on fakebook but hore does.
Hooer
Well, the first option is really good, the "bovine excrement" is hard to pin down without being draconian.
Any variation of a word and it's l33t5p3@k counterparts can be easily detected with regular expressions or regex as we call them in the computing world.
This is some example code that shows how this can happen, using python:
This imports the 'regular expression library in python; basically makes it available to us for use as pre-written code in the current code we want to run.
Let's use a list of words. This is how a list is created in python [ ] mean list, and each of the entries needs to be quoted, single or double, and comma separated, this is so the compiler (what takes code like this and makes it machine code) can understand it.
This is the secret sauce here. Basically we have other FASTER methods to go through a set of characters, but for this simple example, we have defined a key:value pair. Kind of like firstname: John, lastname: Doe. This construct leetspeak_dict is a dictionary, and they just contain the k:v pairs. So when 'a' is encountered, later on we will check 'a' against a capital A, @, and 4, which are all common substitutions.
This part just pre-compiles the patterns for using a for loop for each character in the words from the word_list that we defined earlier.
You could test this using re.match, perhaps even re.fullmatch, but they are fiddly at times. This is just to illustrate.
TL;DR: Your first approach makes pinning down your speech much, much harder. Why? I have to filter so many words that regular speech breaks down, and it it gets exponentially more expensive to police your speech.
Character substitution is easy to detect and block. Huge word lists introduce overhead in checking for disallowed words, and it's even worse with phrases in doing a version of the above, typically in something like javascript. Python is a lot easier to understand, so I did python here. You make them pay a lot more money to censor phrases too. So much so that eventually it becomes too much. If everyone does this, the only choice is manual content moderation, which in the event of bans, drives revenue off the platform, and costs them actual money. Even AI cannot reliably do this yet, because we change the euphemisms constantly. The Euphemistic Treadmill, that liberals un-ironically use to police speech (aka Retard/Austist -> Neuro-Divergent Person) becomes the reason that they, and AI can't control it. Conservative views are worth money too.
Not only that. You now make normal conversations subject to scrutiny, which hurts the credibility of the platform with normie users. False positives will pee off normies.
Regular expression matching is exponentially harder and more expensive with additional phrases, and creates false positives that will be impossible to police without affecting normies.
WEW LAD! How many cups of coffee did you drink this mornin fren?
While I appreciate your passion, a quick 1 minute search would have led you to this document that shows the hype in OP's video is much ado about nothing.
Check it out:
https://www.t-mobile.com/support/public-files/attachments/T-Mobile%20Code%20of%20Conduct.pdf
Wow and boy did you call it! I had like 3 cups in one GIANT thermos this morning!
I saw it was about section 5 last night, that targets campaigns and all, but I see the writing on the wall, after this it's a small text update to remove campaign and were all on a terror watchlist for calling something obviously gay fruity.
I'm paying them 297.00 per month atm. So I am just ready to get down to 40 per month with an MVNO (I work in the industry and know better, but have had to have the Tmo home internet in the area for my brother, so paying out the nose)
Or make to new words. Adapt
You might just be the one to start a new term that goes viral.