AKMA’s Random Thoughts

One Note

I acknowledge that by posting this note, I repeat and perpetuate my fixation on one as-yet-unimplemented feature of the major search engines. Nonetheless, this morning I encountered another setting in which I would love to be able to run a seeded search.

For those who haven’t heard this rant before, “seeded search” refers to a Google-like prioritized search based on the number and quality of a site’s links — but that the search begins outward from a limited set of touchstone sites (or reference terms). Thus, a seeded search for "biblical hermeneutics" that started from my site would turn up postmodern sympathizers much sooner than it would turn up inerrantists. It’s a moderately simple (not simple with regard to the computational resources involved, I would guess) way of making search results more useful. (It would work especially well with Kevin Marks’s link votes system, I should add.)

I return to this refrain of mine this morning as I wandered over to AudioScrobbler’s Top Artists charts. The charts at AudioScrobbler show strong persistence from week to week; AS users really like Radiohead, Green Day, Coldplay, and Nirvana. That’s fine, and it’s useful information. But what I want to know is, “What do the people who never listen to Rammstein like?” — because that’s likely to be much more useful to me. I’ve never voluntarily listened to Rammstein, and doubt that I’ll be won over any too soon; but if someone likes R.E.M. and [something else], I might be interested to know what the other thing is.

So my new seeded-search request would be for “top artist” and “top song” lists based on one or two seeded elements: the most popular artists among people who like Modest Mouse but never listen to Metallica, for instance, or people who listen to both Philip Glass and Sleater-Kinney, or (better yet) a three-term search for artists preferred by people who like Philip Glass and Sleater-Kinney, but who wouldn’t go anywhere near Jamiroquai. You get the idea.

I realize that the automated suggestion and affinity searches do some of this work; I’m asking for slightly higher degree of client-side involvement. There’s a lot to be found out, that we won’t learn unless we let users figure out the terms of their own queries. (As Jenna’s example illustrates powerfully!)

DRMA (Dave Rogers, who’s actually posting again occasionally, Music Alert): Spirit In The Sky by Norman Greenbaum; Who Do You Love? by George Thorogood; Garden of Simple by Ani DiFranco; Kalifornia by Fatboy Slim; The Sheltering Sky by King Crimson; Look at Me by John Lennon; We'll Deal With You Later by the Beautiful South.

Posted by AKMA at November 1, 2004 10:51 AM | TrackBack