Adaptive Personal Information Filtering System that Organizes Personal Profiles Automatically

Abstract

An ada ptive personal information filt ering system is prese nted. Th e ad a ptive lea rning alg orithm of the syst em is derived from th e statisti cal consid e ration - what keywords are useful for retrieving items in whi ch th e user want to see . Ac cording to th e algorithm , th e system orga nizes a personal profile to re fl e ct th e use r's intere sts from an e mpty profile through usersys tem inte ra ction - th e system dis plays items a nd th e user returns th e answer , " int eresting" or " unint ere sting " . Th e syst em gives the user items according to his or her interest e stimated using th e user 's personal profil e. Th e density of th e interesting items at th e he ad of the sequ en ce of items sort ed by th e s yst em is as high as that in th e items re tri eved using the keyword mat ching method whose Boole an expression of keyword s was mad e by a rese ar cher who has mu ch expe ri ence in pat e nt retrieval. Th e number of th e intere sting items included th e head of th e so rt ed sequ ence is larger th a n that in th e retrieved items. A fi eld test is exe cut ed to ensure that th e system respond s to th e varied in terests of many users. Th e items used in the test a re Japanes e newspap er h ea dlines which are s upplied at a rat e of 600 per day from info rm ation providers. Not only researchers and eng in ee rs in various tec hni cal fi elds , but also s ta ff members participat ed in th e te st as users. Th e results of th e te st sugge st th a t th e system ca n be put to pra cti cal use.