[Wikipedia-l] Automatic vandalism detection?

[Guillaume Blanchard]

> On 4/11/03 10:46 AM, "Marco Krohn" <marco.krohn at gmx.de> wrote:
>
> > Hi all,
> >
> > I saw just by chance that [[Rust]] was vandalized by simply adding the
word
> > "bull shit" to the text. Thinking about this I asked myself if it
wouldn't be
> > possible to add an automatic detection for this kind of vandalism?
Something
> > in the way the spam detection programs work. For the beginning a very
simple
> > heuristic algorithm that detects certain word  / word combinations would
be
> > sufficient.
> >
> > Of course the algorithm should not block the modification, simply
because it
> > is impossible to prevent false positives, but it could add the page to a
new
> > special page, making it easier to detect obvious cases of vandalism.
> >
> > Being still quite new to wikipedia I am not sure if this fits well in
the
> > wikipedia framework and if it is realizable, but nevertheless I would
like to
> > hear your comments.
> >
> Without calling it "vandalism", a word I still think is inappropriate for
> Wikipedia, automatic sifting of contributions is certainly a good idea.
>
> It would indeed be very interesting to see what would be the result of
> Bayesian ranking of contributions.

We may also detect big article size decrease (ex: 1 Kb -> 20 bytes).
I think the easer way to show hypothetical "vandalism" is to use different
font (color, italic or font) in RecentChange.
For exemple, display "red" the article where someone just added words like
"fuck" or big size decrease can be very useful.
It can be an option in Preferences.

Aoineko