You could read the text into a scalar and regex the &'s out. See Re^4: Why XML not well formed? for a regex that would work. Granted, if this is a large document, this would probably not be the answer.