ID of anonymous posters via word analysis?

[Matthew J Ghio]

Curtis D. Frye <cfrye at ciis.mitre.org> wrote:

> True, though the probability that two individuals would (over)use a
> particular word or phrase is high enough where "heating things up"
> would be unjustified, especially if spoofing were involved.  Consider
> the reverse of the analytical process -- I want everyone to believe I'm
> Joe X, so I do a text analysis of his messages, write my own, analyze
> my message in comparison with Joe's, and modify it until the (or an)
> engine's algorithms spit out a score indicating that I'm Joe.  Spoofing
> deluxe!
>
> I don't mean to say that informal analysis doesn't have its place, but
> we need to be careful about jumping to conclusions and potentially
> "heating things up" for innocent individuals or "convicting" them in
> the Court of Net.Opinion absent sufficient proof.  I would agree that
> these analyses might form the basis for a reasonable suspicion that a
> particular individual is resposnible for bothersome anonymous posts,
> providing grounds for sysadmin notification.

You can use electronic equipment to disguise your voice on the phone
too, it's just not particularily easy or convienient to do.  Nobody's
saying that this would be convicting evidence, it's like testifying that
you recognized the voice of a caller on the phone.  (Altho electronic
analysis of anonymous callers voices have been used as evidence in
court.)