Rather than extracting the text, could you just render a cached copy of the full web page? I don't see how anyone would have a problem with that. Many times you will end up missing parts of the article anyway if you try to pare it down.