- From: Paul Francis <francis@cactus.slab.ntt.jp>
- Date: Fri, 23 Jun 95 09:53:33 JST
- To: rating@junction.net, uri@bunyip.com, www-talk@www10.w3.org
- Cc: wex@media.mit.edu
Regarding the labeling of information, I just today was made aware, through the bounce-digerati@ai.mit.edu mailing list, of a service at MIT's Media lab that helps one find things on the Web that they may like. To cut-and-paste from a discussion on bounce-digerati@ai.mit.edu: From: medlar@ua.com (Art Medlar) Subject: Audience Discovery Sender: bounce-digerati@ai.mit.edu For a truly excellent example of a potential technical solution to a part of this problem, have a look at: HOMR -- the Helpful Online Music Recommendation Service http://homr.www.media.mit.edu/projects/homr/ The system collects your ratings for a number of music performers, does some magic statistical correlations with other people's ratings, and returns a list of other music that it determines you might like, based on the likes and dislikes of other users who seem to share your taste. From: Alan Wexelblat <wex@media.mit.edu> Since no one else has commented and since HOMR originated in our group, I should note that there's another project using similar technology, called Webhound, which helps people discover Web pages of possible interest. http://webhound.www.media.mit.edu/projects/webhound You can find these projects, their authors and other Agents Group people, and more work (and papers about same) off the Agents group page: http://agents.www.media.mit/groups/agents I copied Alan Wexelblat so that perhaps he could comment. Alan, for your background, this discussion was initiated over the concern of government regulations on the available to minors of adult-only material on the Web, and on what technology could be used to label and filter such material. It occurs to me that an automatic-grading system through correlation of peoples' opinions like you have built could just as easily be used to filter out unwanted resources as to "filter in" wanted resources. Perhaps requested resources (or, their URLs) could first be passed through a grading system such as yours, and automatically labeled according to correlation of peoples' taste and previous recommendations on the labeling or the resource. Comments? PF
Received on Thursday, 22 June 1995 20:54:27 UTC