From: Eugene Leitl (eugene.leitl@lrz.uni-muenchen.de)
Date: Wed Jun 14 2000 - 12:55:00 PDT
Karl Anderson writes:
> Make information filter building part of information consumption.
> Make it near-effortless and rewarding. Make the trails that we're all
> following in the info wilderness (sorry for the stupid metaphor)
> shared, instead of isolated, and make it easy to find the paths of
> others who share affinites, that should be a start.
Information filtering has to be adaptive, using feedback from the
user. There should be a clustering algorithm, creating groups of users
according to common interests. This way, you don't have to carve out
all of your profile by hand, just adjust the clustering threshold.
Information filtering in a push medium should be both person-based
(easiest, just use a ranking matrix of user vs. user, and attach
voting urls to the end of each message posted by user, one to
increment the rank, one to decrement the rank), and content-based,
which will probably require some neuronal net of a nontrivial size
(=too much crunch for today's machines).
This archive was generated by hypermail 2b29 : Wed Jun 14 2000 - 13:59:52 PDT