Also see Text::Unidecode and especially for sanitizing titles for URLs, Text::CleanFragment.

Both err rather on the side of leaving things out rather than keeping things in.

It seems your regular expressions attempt to remove whole Unicode character planes. Personally, I would explicitly allow some character planes or look at the unicode properties (maybe via Unicode::Tussle to find out whether a character is part of a script.

Also consider what you want to do with character art: (╯°□°）╯︵ ┻━┻