For web scraping I used htmltidy (https://en.wikipedia.org/wiki/HTML_Tidy) which cleaned it sufficiently that I could run XSLT over it (gags at the memory)