On 28 February 2010 14:31, Gerard Meijssen <gerard.meijssen(a)gmail.com> wrote:
Hoi,
I came across this in my reader.. I think it is of general interest..
Thanks,
GerardM
http://news.slashdot.org/story/10/02/28/1946216/Developing-a-Vandalism-Dete…
While I, of course, appreciate anyone wanting to help deal with
vandalism, it seems the person responsible for this underestimates the
efficacy of our existing methods (automated, semi-automated and
manual). This paper by the same person (he links to it in a slashdot
comment):
http://www.uni-weimar.de/medien/webis/publications/downloads/papers/stein_2…
is rather interesting. It contains some good statistics. He is of the
opinion, however, that we should prefer what he called "high-recall"
(few false negatives) over "high-precision" (few false positives)
since it avoids readers seeing vandalised versions. I disagree, since
he hasn't taken into account how annoying it is for editors to be
reverted by a bot and the harm caused by losing those editors.
FlaggedRevs (which I remain ever optimistic about!) will provide a
much better solution.