Often times people will write their content in Microsoft Word and then cut/paste their work into a website. By default microsoft uses their own character encoding and they use control characters. These control characters are not used in other character encodings and will not properly show up when redisplayed to the user in html.

This is not a very large problem at this time.

There are only 3 datasets that have the old ansii based control characters

and

only 46 datasets that have the new microsoft unicode characters.

That is NOT bad for ~3000 datasets.