Hi, Perhaps it might be interesting to allow the first person to submit a URL to predict who will likely be interested in that URL.
For example, when submitting a URL for vim, I might indicate that people who bookmarked a recent but popular Windows related URL would likely not like this link. But someone who bookmarked a recent but popular python URL and did not bookmark a recent but popular emacs URL would be. These predictions would be used to give URL recommendations and combat spam. Predictions that turn out to be accurate (by determining whether the URL was bookmarked or not for various prediction rules) would be used to reward the submitter in some way. Of course, the submitted URL could be recommended more aggressively based on these rules if it turns out that the prediction rules are turning out to be accurate based on initial data. A variation on this idea is to allow multiple people to make predictions, not just the first submitter of a url. This proposal essentially turns collaborative filtering into a prediction game. Why not just use an automated algorithm? There is a lot of creativity involved in coming up with prediction rules. The search space is huge. Moreover, it might be fun to see how well your prediction rules work out. Amir _______________________________________________ discuss mailing list [email protected] http://lists.del.icio.us/cgi-bin/mailman/listinfo/discuss
