It appears to be (badly) attempting to return first some tag after an H1 tag (skipping all tags following the H1) in $1, then the longest sequence of word characters immediately following those tags in $2.