I'm not sure about the LaTeX generation side, but you could start with the HTML::Parser module, described in the latest issue of The Perl Journal, to parse through the html document's structure.