Appeal 2007-1122
Application 09/966,414
42 based on the ratings of the other subscribers in the group. (Col. 8,
ll. 55-58; Fig. 6.)
4. Payton teaches that a collaborative filtering system 42 synthesizes the
subscriber profiles 40, predicts which of the available items 36 each
subscriber may be interested in or may request, and produces a list 44
of recommended items for each subscriber. (Col. 5, ll. 12-16.) The
list 44 may include items that a particular subscriber has never
previously requested. (Col. 5, ll. 16-20.)
5. A scheduling processor 46 periodically receives an updated list 44 of
recommended items for the subscriber from the collaborative filter 42
(step 68). (Col. 6, ll. 63-67; Fig. 3a.) The scheduling processor 46
transmits the changes in the lists to the subscribers (step 70) and
merges the new additions to the list into a refresh queue (step 72).
(Col. 6, l. 67 to col. 7, l. 4; Fig. 3a.)
6. The scheduling processor 46 retrieves an item from the refresh queue
47 (step 90), and updates the subscriber profile 40 to reflect the
storage change that will occur when that item is received by the
subscriber's local server 28 (step 94). (Col. 7, ll. 36-47; Fig. 3c.)
7. In the preferred embodiment, the collaborative filter 42 periodically
re-computes subscriber similarity groups (step 152). (Col. 8, ll. 61-
63; Fig. 7a.) Based on these groupings, the filter 42 predicts ratings
6
Page: Previous 1 2 3 4 5 6 7 8 9 10 11 12 13 Next
Last modified: September 9, 2013