Archive of UserLand's first discussion group, started October 5, 1998.

Re: Collaborative Filter theory.

Author:Paul Snively
Posted:11/10/1999; 7:26:54 AM
Topic:Collab filtering at my.userland.com?
Msg #:12914 (In response to 12908)
Prev/Next:12913 / 12915

-- wrote:

Whether you gather the preference data by query "Did you like it?" or inference "They looked at this page for N seconds." it's still collaborative filtering.

In that case, I'd expect to wind up running afoul of someone's patent.

There's no reason to have a cutoff at all. Sort the data according to predicted interest, and the user can quit reading when they're tired.

If you do that, all you've done is create a search engine where you've tried to predict the search. I do agree that an easy shortcut to seeing all choices is needed; I don't agree that presenting all choices up front is a good idea unless the total number of choices is small--say, no more than seven, which, as I recall, is what cognitive science tells us is the maximum number of things the average human can contemplate at once.

Large quantities of data are more important than clever algorithms at providing useful results.

Two words: "search engines."

Or, if you want more detail, "AltaVista" (large quantities of data) vs. "Google" (clever algorithms, useful results).

Filtering entire lists is far less useful than filtering specific topics within a list.

And I believe I've said as much.

There are responses to this message:

This page was archived on 6/13/2001; 4:53:26 PM.

© Copyright 1998-2001 UserLand Software, Inc.