Still playing with my profanity filter :-)

A common issue is with users trying to use alternate characters in place of recognized characters, trying to get around the filter. Examples including using @ instead of a, $ instead of s, ! or l (lowercase L) instead of i; ! or I (uppercase i) instead of L, and so on.

I have a ton of regex written to filter out profanity, and I duplicate the same things in each one of them:

[s\$]
[a\@]
[il!]

I recently had to deal with a new variation, and have spent most of my day modifying all of the filters to catch it :-/

As a long term fix, I'm thinking of a way to apply this variations at the beginning of the filter, so that I only have to modify them once.

My initial thought is:

1. Create an associative array with all of the potential workarounds; eg, '$' => 's', '@' => 'a', and so on.

2. Split the string by \s

3. Loop through the new array with the associative array, apply the workarounds, then perform a substitution if the new modified word matches a filtered word

Something like:

%badwords = (
  'cat' => 1,
  'dog' => 1
);

%workarounds = (
  '$' => 's',
  '@' => 'a',
  'l' => 'i',
  '!' => 'i',
  'j' => 'i',
  '+' => 't'
);

$text = 'this is a c@+';

@arr = split(' ', $text);

foreach $original (@arr) {
  $_ = $original;

  foreach $key(keys %workarounds) {
    if (s/\Q$key\E/$workarounds{$key}/i && exists($badwords{$_})) {
      $text =~ s/\Q$original\E/****/i;
    }
  }
}

print $text;
# this is a ****

This works, but given a post of 5,000 words and running each word over a ton of regex filters, this would be SUPER slow!

Any other suggestions?