Treat similar looking character as identical when checking blocked words to prevent automodevasion
Treat similar looking character as identical when checking blocked words to prevent automod evasion.
So if the word "Apple" is blocked, the word "Àṗṗĺĕ" would be blocked too.
Blocking every variation of a word would literally take 76 days at 100 terms per 30 seconds (rate limit by Twitch). This is for obvious reasons not a solution. And the current moderation tools are thus obviously insufficiant.
https://twitter.com/thomsimonson/status/1429472208506822659
This feature request is in response to the ongoing hate-raids mainly targeted at LGBTQIA+ people
twitchdobetter
Ysgramor3
shared this idea
-
RealityRipple
commented
Most programming languages have a Culture Invariant string comparison, that treats all accent marks as irrelevant, comparing only the base characters. This could potentially be solved by simply changing a single function call.
-
WhazzItToYa
commented
And now I'm seeing messages come through with character modifiers or diacritical characters sprinkled in. E.g.: "B͟est viewers at ..." which is a regular 'B' followed by unicode "COMBINING DOUBLE LOW LINE" (0x033f). It evades the blocked term 'best viewers' because of the extra character between the B and E.
-
WhazzItToYa
commented
Agreed. There's a certain "buy follows" bot that comes into my channels all the time, using homoglyphs in the message that will never be caught by the blocked terms feature.