How "essentially HTML" is this HTML? If it's actually HTML, I would instead decode it then use HTML::Strip to remove the HTML. Writing a reliable regular expression to parse potentially nested tags properly is difficult.