Perhaps you could run each message through Google translate or something similar (e.g. English -> French -> English) to obfuscate the writing style. Otherwise, as others have already pointed out, it is very indicative of who the author is.
i wonder if you could repeat with different languages and take some kind of semantic median? interesting problem.
[just to clarify, given the replies - i am talking about repeating in parallel, not in series. the idea being that errors might cancel out, rather than accumulate]
I think all conversations should be run back and forth through Google Translate until they reach equilibrium (a la Translation Party). For added awesomeness, do a multi-way equilibrium instead of just two languages.
I (translation of a la carte party) have achieved a balance
between the universal and Google, and I need to translate
the conversation. Not only, Blip.fm, from how to move from
the language of two additional multi balance.
I tried it with your comment, and it came back the same. However, English-->French-->Polish-->English came back with something altered:
Perhaps start every message by the Google Translate or something similar (eg English -> French -> English) to hide the writing style. Otherwise, as already mentioned, this is very revealing of who the author.
You'd need to run it through a non-deterministic model for that to work. Maybe select a couple languages at random to run it through, rather than the same one every time.
