You could build that feature dynamically. When a user tries to use a disallowed word and then uses an allowed word, you could log that word pair. Combine and an

You could build that feature dynamically. When a user tries to use a disallowed word and then uses an allowed word, you could log that word pair. Combine and anonymize those logs and you'd be able to show the most likely replacement words for any word that people often try to use.