Having done this kind of thing a few times, let me suggest that IMHO, this is a case where regex is better than parsing. Ignoring sites continually changing their formats, the main problem I found was that there were a lot of web-sites using broken html, as in it wouldn't validate and thus parsing was problematic.

This is one of those cases where, in theory, parsing is better, but it has to be very forgiving parsing, to the point that I found it unreliable. Disheartening.