improving cpunk S/N via grouplens/filtering
an interesting new field called "collaborative filtering" seems to have strong potential in helping separate signal-to-noise, and might be of interest to some cpunks here as a pet project. Brad Miller is working on his PhD. in this area and put together a nice system inside many news readers that allows people to read/ratings articles and track their ratings. the system uses the heuristic that people who agree or disagree consistently on ratings on particular messages in the past will tend to do so in the future, and it uses this to extrapolate ratings on messages a person has not yet rated. essentially he has all the technical stuff down and mostly just needs trial users. it wouldn't be hard to plug it all into the cpunk list to see how well it works. its the most promising route I know of at the moment, for solving an extremely difficult problem that pervades cyberspace. however, B.M. has been working on the project for a long time and at least from my perspective it doesn't seem to have reached "critical mass" yet, which surprises me. anyway, the URL http://www.cs.umn.edu/Research/GroupLens/grouplens.html is anyone here familiar with this project? what do you think about how effective it is?
"Vladimir" == Vladimir Z Nuri <vznuri@netcom.com> writes:
[GroupLens] Vladimir> http://www.cs.umn.edu/Research/GroupLens/grouplens.html Vladimir> is anyone here familiar with this project? what do you think Vladimir> about how effective it is? I like the idea, but they've never covered anything other than Usenet, and newsgroups I don't read. The little feedback we've gotten after putting it in Gnus indicates either that it is not useful, or that it probably hasn't reached critical mass yet. I proposed GroupLensing this list to Brad early this year, but he never followed up on it. To the luser who unsubscribed last week after complaining about his wimpy mail program, I would add: Gnus runs on Emacs 19.34 which fully supports Microsoft Windows and is available as a precompiled binary, so there is no excuse for using mail and newsreading software on that platform without threading, sorting, kill files, etc. Add in the mailcrypt package and an up-to-date copy of Raph's remailer list, and you have basically point and click access to remailer chaining and PGP. -- steve@miranova.com baur Unsolicited commercial e-mail will be billed at $250/message. What are the last two letters of "doesn't" and "can't"? Coincidence? I think not.
participants (2)
-
Steven L Baur -
Vladimir Z. Nuri