I think rindolf received a TPF grant to write a mediawiki markup parser, maybe he has some results already that you could use.

Other modules like Parse::MediaWikiDump look promising too.

As for your X-Y problem: you could just parse wikipedia's HTML output, there's a myriad of modules for that on CPAN.