You want a proper html 5 parser that can handle non valid documents. And the fastest one is https://github.com/kovidgoyal/html5-parser over 30x faster than html5lib