GROUP COMPOSITION BASED RECOMMENDER SYSTEM AND METHOD

• US 20150186934A1
• Filed: 08/26/2013
• Published: 07/02/2015
• Est. Priority Date: 08/24/2012
• Status: Abandoned Application

• Current Assignee: Nederlandse Organisatie Voor Toegepast-Natuurwetenschappelijk Onderzoek Tno
• Original Assignee: Nederlandse Organisatie Voor Toegepast-Natuurwetenschappelijk Onderzoek Tno
• Inventors: Van Deventer, Mattijs Oskar, De Wit, Joost Jelmer
• Application Number:
  US14/423,644
• Publication Number:
  US 20150186934A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06Q 30/0251 Targeted advertisements
• CPC Class Codes: G06Q 30/0631 Item recommendations
• CPC Class Codes: H04N 21/4668 for recommending content, e.g. movies
• CPC Class Codes: H04N 21/4756 for rating content, e.g. scoring a recommended movie
• CPC Class Codes: H04N 21/4826 using recommendation lists, e.g. of programs or channels sorted out according to their score

Abstract

A recommender system stores rating values for items associated with identified users and groups of users and information representing users that are part of the groups of users. A correlation module computes correlations between a first group and other groups, by retrieving the information representing users in the first group and the other groups from the group composition storage system, retrieving the information representing rating values for items from a reference set associated with the users of the first group and the other groups and computing an aggregate of correlations between the rating values for items from the reference set associated with pairs of users, each pair comprising a respective one of the users of the first group and a respective one of the users form one of the other groups. A composition module con- figured to compute an aggregate of parameter values for the group from parameter values, such as rating values for items, associated with the other groups, wherein the other groups that are used to compute the aggregate of parameter values are selected and/or weighted dependent on the aggregate of correlations. By aggregated correlations as a group based similarity score, the adaptability of the recommender system to group composition is increased.

Claims

1. A recommender system, comprising

a rating storage system;

a group composition storage system;

a correlation module configured to compute correlations between a first group and further groups, by

retrieving information representing users in the further groups from the group composition storage system,

retrieving information representing rating values for items from a reference set associated with the users of the first group and of the further groups indicated by the information representing users in the further groups;

computing correlations between the rating values in the retrieved information representing the rating values for the items from the reference set for items from the reference set associated with pairs of users, each pair comprising a respective one of the users of the first group and a respective one of the users form one of the further groups;

computing an aggregate of the correlations for respective ones of the further groups;

a composition module configured to compute an aggregate of parameter values for the group, from parameter values associated with the other groups, wherein the further groups that are used to compute the aggregate of parameter values are selected and/or weighted dependent on the aggregate of correlations for the respective ones of the other further groups.

2. A recommender system according to claim 1, wherein the composition module is configured to select a set of the further groups for use to compute the aggregate of parameter values, to retrieve the parameter values selectively for the further groups in the selected set and to compute the aggregate of parameter values from the retrieved parameter values.

3. A recommender system according to claim 1, wherein the correlation module is configured to compute a plurality of alternative aggregates of correlations, each for a different sets of pairs, different sets of pairs combining the users of the first group differently with users form one of the further groups in the pairs, the score value composition module being configured to compute the aggregate of rating values selected and/or weighted dependent on a maximum of the alternative aggregates of correlations.

4. A recommender system according to claim 1, wherein the composition module is a score composition module for computing a collaborative rating score, the parameter values being rating values for a further item, the score composition module being configured to compute an aggregate of rating values from rating values for the further item associated with the further groups, wherein the further groups that are used to compute the aggregate of rating values are selected and/or weighted dependent on the aggregate of correlations.

5. A recommender system according to claim 4, having an interface for coupling to a user terminal, the score value composition module being configured to compute said aggregate of rating values for a plurality of further items, the recommender system comprising an item selection module configured to make a selection from the plurality of the items based on the aggregate of rating values computed for the plurality of the further items and to transmit an indication of the selection to the user terminal, together with addresses for access to the selected further items from a data store system.

6. A recommender system according to claim 1, comprising retrieving information representing users in the first group from the group composition storage system in combination with retrieving the information representing users in the further groups from the group composition storage system.

7. A recommender system according to claim 1, comprising receiving identifiers of the users in the first group from a user terminal based on input that identifies the users at the user terminal.

8. An electronic data access system, comprising the recommender system according to claim 5, wherein the score composition module is configured to compute aggregates of rating values for a plurality of further items from rating values for the further items associated with the further groups, the electronic data access system comprising an item selection module configured to select a subset from the further items based on the aggregates of rating values, the recommender system being configured to transmits information indicating the items in the selected subset to a user terminal, with information that enables addressing the selected items for retrieval from an item storage system.

9. An electronic data access system, comprising the recommender system according to claim 5 and a data storage system for storing the items, the data storage system comprising an interface unit configured to supply an item to the user terminal in response to transmission of the address from the user terminal to the data storage system.

10. A machine implemented method of generating Recommendations, wherein processing a request for a recommendation for a first group of users comprises

a computer system retrieving first information representing in further groups from a group composition storage system,

the computer system retrieving second information representing rating values for items from a reference set associated with the users of the first group and the further groups based on the first information;

the computer system computing correlations between the rating values from the second information for items from the reference set associated with pairs of users based on the retrieved second information, each pair comprising a respective one of the users of the first group and a respective one of the users from one of the further groups;

the computer system computing an aggregate of the correlations for respective ones of the further groups;

the computer system computing an aggregate of parameter values for the first group, from parameter values associated with the further groups, wherein the further groups that are used to compute the aggregate of parameter values are selected and/or weighted dependent on the aggregate of correlations.

11. A method according to claim 10, comprising

computing a plurality of alternative aggregates of correlation, each for a different sets of pairs, different sets of pairs combining the users of the first group differently with users form one of the further groups in the pairs,

computing the aggregate of rating values being selected and/or weighted dependent on a maximum of the alternative aggregates of correlations.

12. A method according to claim 10, comprising computing an aggregate of rating values for a further item from rating values for the further item associated with the further groups, wherein the further groups that are used to compute the aggregate of rating values are selected and/or weighted dependent on the aggregate of correlations.

13. A method according to claim 12, comprising

computing said aggregate of rating values for a plurality of further items,

making a selection from the plurality of the items based on the aggregate of rating values computed for the plurality of the further items and

transmitting an indication of the selection to a user terminal, together with addresses for access to the selected further items from a data store system.

14. A tangible computer readable medium, comprising a program of instructions for a programmable computer system that, when executed by the programmable computer system will cause the programmable computer system to process a request for a recommendation for a first group of users by executing the steps of

retrieving first information representing from a group composition storage system, the first information representing further groups, other than said first group;

retrieving second information representing rating values for items from a reference set associated with the users of the first group and the further groups based on the first information;

computing correlations between the rating values from the second information for items from the reference set associated with pairs of users based on the retrieved second information, each pair comprising a respective one of the users of the first group and a respective one of the users from one of the further groups;

computing an aggregate of the correlations for respective ones of the further groups;

computing an aggregate of parameter values for the first group, from parameter values associated with the further groups, wherein the further groups that are used to compute the aggregate of parameter values are selected and/or weighted dependent on the aggregate of correlations.