"a lot of people copy and paste directly from Word... "

GrandFather's suggestion that you use a whitelist and tr/// to remove anything NOT on the whitelist is a decent solution, assuming you don't have any need for chars outside printable ASCII.

An alternate -- draconian but even less INsecure (all other things being equal) -- is to flat out reject any entry which contains (for example) MS Word artifacts -- smart quotes, and so forth -- and return the poster to the data entry form with an explanation that your form accepts only text. (Posters need not be excessively inconvenienced: any M$ Word doc can be saved as text within Word itself and when copy-pasted from a text version should contain none of the chars which "screw up (your) RSS feeds."