Interactive hybrid recommender system
First Claim
Patent Images
1. A hybrid recommender system for delivering content that fits a user preferences, implementing a process comprising the steps of:
- a. manually defining one or more initial stereotypes by an expert;
b. creating an affinity vector of stereotypes relating to each specific user who registers onto the system, thereby creating a specific profile for each user, wherein said affinity vector for each user is created according to her response to a set of questions, where each question compares several representative items, and a form, based on a decision tree and specifying said user'"'"'s interests, and wherein said affinity vector is saved in a database;
c. generating recommendations for a specific user according to said initial stereotype and said affinity vector of stereotypes, wherein said recommendations are generated using a matching engine that computes the relevance of an item to the user, thereby creating a recommendation list of items to incorporate into the database;
d. receiving feedback from user regarding specific items picked by him, wherein said feedback can be either positive or negative; and
e. updating said affinity vector of stereotypes and said recommendation list of items saving them in said database;
wherein the content delivered to the user contains a predefined number of the most relevant items to her according to said recommendation list.
4 Assignments
0 Petitions
Accused Products
Abstract
A hybrid recommender system, in which the initial stereotype is manually defined by an expert and an affinity vector of stereotypes relating to each specific user who registers onto the system, is created to define a specific profile for each user. Recommendations for a specific user are generated according to the initial stereotype and the affinity vector of stereotypes. A binary feedback, from user regarding specific items picked by him is received (e.g., while of the item), which can be either positive or negative. Then the affinity vector of stereotypes is updated.
-
Citations
18 Claims
-
1. A hybrid recommender system for delivering content that fits a user preferences, implementing a process comprising the steps of:
-
a. manually defining one or more initial stereotypes by an expert; b. creating an affinity vector of stereotypes relating to each specific user who registers onto the system, thereby creating a specific profile for each user, wherein said affinity vector for each user is created according to her response to a set of questions, where each question compares several representative items, and a form, based on a decision tree and specifying said user'"'"'s interests, and wherein said affinity vector is saved in a database; c. generating recommendations for a specific user according to said initial stereotype and said affinity vector of stereotypes, wherein said recommendations are generated using a matching engine that computes the relevance of an item to the user, thereby creating a recommendation list of items to incorporate into the database; d. receiving feedback from user regarding specific items picked by him, wherein said feedback can be either positive or negative; and e. updating said affinity vector of stereotypes and said recommendation list of items saving them in said database; wherein the content delivered to the user contains a predefined number of the most relevant items to her according to said recommendation list. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18)
-
Specification