| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide an unpredictable recommendation matched to a user's interest for an article for a daily use or a broadcast content. <P>SOLUTION: The information recommendation system comprises a taste model formation part 45 which learns and stores a user's taste based on a user's action record; an evaluation list formation part 46 which calculates a selection probability for a plurality of recommendation candidates based on the taste model and forms an evaluation list showing a probability of user's selection for each recommendation candidate; a default prediction part 47 which specifies a user's habitual property based on information for the user's action record and determines default prediction values for the plurality of recommendation candidates; and an unpredictability evaluation value calculation part 48 which determines an unpredictability evaluation value showing a degree of unpredictability with a larger difference between the selection probability and the default prediction value which is considered to be higher in the degree of unpredictability. <P>COPYRIGHT: (C)2009,JPO&INPIT |
