Browse Prior Art Database

Personalized Information Navigator

IP.com Disclosure Number: IPCOM000117993D
Original Publication Date: 1996-Aug-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 4 page(s) / 190K

Publishing Venue

IBM

Related People

Nomiyama, H: AUTHOR [+3]

Abstract

Disclosed is a device for saving personal interests used for personalized information filtering.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 29% of the total text.

Personalized Information Navigator

      Disclosed is a device for saving personal interests used for
personalized information filtering.

Memory Model - A tree structure whose nodes are named after their
categories is used to store users' preferences.  Each node in a
static structure of a memory has a set of memory cells.  Each memory
cell contains a keyword, its degree of activation, and a history of
its activation.  The degree of activation L (-1 <= L <= 1) represents
how interested users are in a keyword.  A value of 1 represents a
user's highest preference, while a value of -1 represents the highest
non-preference.  A degree of activation is updated by stimuli created
from the user's operation.  The higher the degree of activation, the
stronger the user's interest.  A history of activation contains
degrees of activation and their time stamps during learning.  So that
the memory model can be used to calculate the level of a user's
interest, active memory, which is a related part of the user's
current interest, is created from the whole memory structure (a
potential memory).  An active memory is created by collecting memory
cells close to a (current focus node).  A current focus node means a
node on which a user is focusing in the potential memory.  More than
one node can be specified.  If the number of the current focus nodes
is greater than 1, each node is weighed with a value Wn (0 < Wn < 1).

      To create an active memory, the degrees of relationship R
(0 <= R <= 1) between the current focus node and all other nodes are
first calculated.  The degrees of relationship between two nodes at a
distance of 1 is defined as R = C/T, where T is the total of the
absolute values of the degrees of activation of memory cells in the
current focus  node, and C is the total of the absolute values of the
degrees of activation of memory cells in the current focus node whose
keywords exist  in the target node and whose signs (plus or minus)
are identical with those in the current focus node.

      The degree of relationship between the current focus node and
the target node is first calculated, and multiplied by all the
degrees of relationship of nodes between the current focus node and
the target node.  If a degree of relationship is less than a
pre-defined threshold  value, the degrees of relationship for all
nodes further from the current  focus node become 0.  The degrees of
relationship for all nodes are calculated.  Next, all the degrees of
activation in memory cells in the  node whose degrees of relationship
with the current focus node are greater than 0 are multiplied by the
degrees of relationship.  If the total number of current focus nodes
is greater than 1, the degrees of activation are multiplied by the
sum of the degrees of relationship for  all the focus nodes and by
the weight Wn.  All the memory cells whose degrees of activation are
greater than 0 are collected and ordered in descending order of th...