Some old version of htmltidy had a handy word-2000 option:

word-2000

Type: Boolean
Default: no
Example: y/n, yes/no, t/f, true/false, 1/0

This option specifies if Tidy should go to great pains to strip out all the surplus stuff Microsoft Word 2000 inserts when you save Word documents as "Web pages". Doesn’t handle embedded images or VML. You should consider using Word’s "Save As: Web Page, Filtered".

I've used it some years ago, with sufficient (but not very pretty) results.

Alexander