Re: author identifiability

From: Spike Jones (spike66@ibm.net)
Date: Sun Aug 20 2000 - 16:38:41 MDT


> >> John Calvin wrote: Gee Spike, isn't handwriting admissable as evidence?
> >
> >It is. We have none.
>
> I used the handwriting example to bring up the point that there are many ways to identify individuals that, while not nearly as "certain" as DNA testing are reliable enough for use as evidence in a Court of Law. How much text would be required to state that duplication of style would be say 1:1000 chance?

Good question. There were some intruiguing stylistic points in the offending
text, such as lions having pups, etc. The others I will not post, for they may
be useful in case the same poster hits again under a different name.

> Could even that level of assurance be accomplished? Will text analysis ever be used by FOrensic scientists in the future?

Hmmm. I guess it depends on if the case is tried in criminal court
or civil.

> >> With a fairly substantial test run would anyone be able to demonstrate the reliability of such pattern recognition software enough to clearly show that even though anyone could have used that particular machine and account they could not have so exactly duplicated style of some named individual? Is that doable?

Ja, but I need to explain what my software does. The user reads in
a bunch of text, then software graphs a blip each time a new word
is used. If you have a book written all by the same person, there is
a lot of new words up front (all of them in fact), then the appearance
of new words gets exponentially less frequent. If a new author's text
is appended, a cluster of new words shows up. Its really a pretty
simple minded program.

I was in theology school when I wrote this. My class mates were
enthralled, as computers had seen little use in that field. The
software cares not which language is used, so we loaded the
Hebrew Pentateuch, and were able to see why the common
theory is that the first five books of the bible were probably
written by at least 4 authors.

Nowthen, the real clues are more stylistic, and even with
advanced tools, there is plenty of guesswork and opinion.
Even today, there is lively debate over whether all the works
attributed to Shakespeare were his, or do they include some
work by Francis Bacon, et al.

Of course, if we have all that other evidence you sited, we
could make a damn strong case. I was not aware that
libraries had cameras, but if they did I would step up to
the plate with literary and stylistic analysis. Now, for such
a tiny piece of text, one page, we neednt use the software,
which would likely be useless for that anyway, but compare
the offending text with the other four posts by you-know-who.
It is clear enough it is the same person.

Nowthen, no one has mentioned the fact that if we attempted
either a criminal case or civil lawsuit on the basis of internet
slander, we are plowing new legal ground. Legals, is this
correct? I know of no one who has successfully prosecuted
for internet slander, even tho I have seen posted many
things that would be illegal pretty much in any other medium.
We might be setting ourselves up for an expensive case
that would go nowhere anyway. Of course I could be
wrong. I hope Greg Burch offers an opinion here.

John, The fact that spam is still legal and so many obvious con
games are being fearlessly played out on the internet makes
me think the chances of prosecution are slim. Nowthen, if
we had solid evidence, such as recorded video of a perp
slinking into the library, doing the classic Snidely Whiplash look-
both-ways- and nyah-ah-ahhhh etc, with time stamps and
traceability to that terminal etc, making a slanderous post,
then we have a hell of a case. But I honestly feel literary
style analysis will not get us there, either in criminal or
civil court. Furthermore, I suspect we will hear nothing
more from you-know-who, but will ever have this ugly
incident as the electronic version of a drive-by shooting.
These are some important and intriguing questions you
raise however.

spike



This archive was generated by hypermail 2b29 : Mon Oct 02 2000 - 17:36:20 MDT