On Oct 30, 2007 11:18 AM, Robert Rohde <rarohde(a)gmail.com> wrote:
If you ignore the substantial problem of text
migrating from one article to
another (admittedly a big deal), then I would have to disagree that it can't
be done automatically.
You and a zillion other people *say* this. But the results never show.
It shouldn't be done in real time, but a computer
with access to the full revision history could figure out how much of the
current text each historical contributor was responsible for, and this could
be used to identify primary authors and add their names via a bot. It's not
a trivial problem but the existing research efforts on the persistance and
evolution of text in wiki articles already largely address these issues.
Thus far the results I've seen are confused by moderately complicated
compound changes which don't confuse human reviewers.
It's not too hard to produce an analysis which can measure roughly how
long a given exact substring has been around. But failing when text is
moved and merged/edited at the same time isn't sufficient for
attribution.
I'd love to see it.. but I haven't yet. Show me the code.
Kwan Ting Chan <ktc(a)ktchan.info> wrote:
No? Software change such that from a certain point on
any edit will
automatically update a credit page. At the same time, start running
something that goes through all the non-deleted edits and update the
page as well.
Yes, we could just extract the long list of names (including offensive
vandal names) from the history, store it, and update it.
What real advantage would this serve over just directing people to the
history page? Lack of duplication doesn't seem like enough of a win
to justify keeping the extra data.
Most of the people who edit a page aren't authors, certantly not in
the copyright bearing sense, and it's unfortunate and confusing that
"Throbbing Monster Cock" gets equal attribution for adding the word
"the" to an article as an actual author.