I guess you could skirt around this by using something to tag the various parts of speech in your original text (using something like Python's NLTK) and replace them with randomly picked synonyms from a thesaurus?

Pretty sure it would obscure the original writer although possibly at the cost of obscuring the original meaning.