If it was a normal XSLT question, I would suggest you to use something like:
But unfortunately the Filtering XSLT seems to be called by something and not simply applied to the field. Using the above example would simply clean out all content of the field (at least that is what it does for me).
So the best thing I could come up with was to filter out all unwanted HTML tags, using something silly like this:
<stylesheet xmlns="http://www.w3.org/1999/XSL/Transform" version="1.0">
<output omit-xml-declaration="yes" method="xml" cdata-section-elements="script"></output>
<template match="/ | node()">
<template match="a | b | br | del | div | em | font | h1 | h2 | h3 | h4 | h5 | h6 | i | img | li | ol | p | strong | span | sup | table | td | th | tr | ul">
Now this is no where near complete, it misses all kinds of HTML tags not even speaking of uppercase variants.
So I would have to conclude that the answer given by Siva is your best option.