Apparatus and method for recommending to an individual selective information contained within a computer network

• US 6,606,624 B1
• Filed: 12/20/1999
• Issued: 08/12/2003
• Est. Priority Date: 08/13/1999
• Status: Expired due to Term

• Current Assignee: Regents of the University of California (University of California)
• Original Assignee: Regents of the University of California (University of California)
• Inventors: Goldberg, Kenneth Y.
• Primary Examiner(s): Metjahic, Safet
• Assistant Examiner(s): CHEN, TE Y
• Application Number:
  US09/467,447
• Time in Patent Office: 1,331 Days
• Field of Search: 707/10-12, 707/3, 707/5, 707/6, 709/200-219, 706/45-56, 706/16-18, 706/25-26, 704/256, 382/211-219, 395/2-2-, 705/27
• US Class Current: 1/1
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 16/9536 Search customisation based on social or collaborative filtering
• CPC Class Codes: G06F 16/9538 Presentation of query results
• CPC Class Codes: Y10S 707/99936 Pattern matching access

Abstract

The method of the invention is executed by a computer under the control of a program stored in computer memory. The method includes the step of accumulating preference data for a set of individuals. The preference data is transformed from multi-dimensional data to lower-dimensional data using a principal component analysis. The lower-dimensional data is then converted into a recommendation map where predicted preference information for a selected individual is easily computed. Recommended content is then routed to the selected individual. The invention has been tested, by way of example, to recommend jokes to users based on their ratings of sample jokes.

Claims

1. A method to recommend items to at least one user based upon data collected from N users where N>

>

1, the method executed by a computer under control of a program, the computer including a memory storing the program, the method comprising the following steps;

(a) during an initial on-line phase, accumulating a database of preference patterns based upon input from N users, during which phase no recommendations are made;

(b) during an initial off-line phase, processing the database accumulated at step (a) to identify a predictor set of items for all future users to rate and to transform high-dimensional preference patterns to lower-dimensional data, and converting said lower-dimensional data into a recommendation map;

(c) during a second on-line phase;

(i) collecting high-dimensional preference data for said new user; and

(ii) generating recommendations for said new user based on data collected in step (c)(i) and based upon said recommendation map derived in one of step (b) and step (d);

wherein time-consuming processing steps associated with step (b) and step (d) are avoided and effective magnitude of N is reduced; and

(d) during subsequent off-line phases;

processing said database of preference patterns based upon input from more than N users accumulated from step (a) and step (c)(i);

transforming high-dimensional preference patterns to lower-dimensional data; and

converting said lower-dimensional data into a recommendation map.

2. The method of claim 1, where step (a) includes collecting high-dimensional preference data on many items from a set of users.

3. The method of claim 1, wherein step (b) includes:

coordinating preference information for a selected user; and

mapping said preference information to said recommendation map.

4. The method of claim 3, further comprising the steps of:

routing chosen content from a set of deliverable content to said selected user in response to said mapping step.

5. The method of claim 1, wherein step (a) includes:

(i) delivering selective content to users; and

(ii) recording responses of users receiving said selective content to form said preference data.

6. The method of claim 5, wherein step (a)(i) includes delivering said selective content to users through an interactive graphical user interface.

7. The method of claim 5, wherein step (a)(ii) includes recording fine-grained responses of said users receiving said selective content specified through a fine-grained rating menu.

8. The method of claim 1, wherein step (a) includes accumulating a dense matrix of preference data for said set of individuals based on dividing objects to be rated into a predictor set and a recommendation set.

9. The method of claim 8, wherein step (a) includes selecting objects having high variance in their preference values for said prediction set.

10. The method of claim 1, wherein step (a) is performed interactively during said on-line phase.

11. The method of claim 4, wherein said mapping and routing steps are performed in an on-line phase.

12. The method of claim 1, wherein step (b) is performed in an off-line phase.

13. The method of claim 1, wherein, step (b) includes clustering said lower-dimensional data to form said recommendation map.

14. A computer readable memory storing a routine that upon execution directs a computer to function so as to recommend items to at least one user based upon data collected from N users where N>

>

1, the routine upon execution carrying out the following steps;

(a) during an initial on-line phase, accumulating a database of preference patterns based upon input from N users, during which phase no recommendations are made;

(b) during an initial off-line phase, processing the database accumulated at step (a) to identify a predictor set of items for all future users to rate and to transform high-dimensional preference patterns to lower-dimensional data, and converting said lower-dimensional data into a recommendation map;

(c) during a second on-line phase;

(i) collecting high-dimensional preference data for said new user; and

(ii) generating recommendations for said new user based on data collected in step (c)(i) and based upon said recommendation map derived in one of step (b) and step (d);

wherein time-consuming processing steps associated with step (b) and step (d) are avoided and effective magnitude of N is reduced; and

(d) during subsequent off-line phases;

processing said database of preference patterns based upon input from more than N users accumulated from step (a) and step (c)(i);

transforming high-dimensional preference patterns to lower-dimensional data; and

converting said lower-dimensional data into a recommendation map.

15. The computer readable memory of claim 14, further comprising a recommendation module to coordinate preference information for a selected individual and map said preference information to said recommendation map.

16. The computer readable memory of claim 15, further comprising a content delivery module to route said selected individual chosen content from a set of deliverable content.

17. The computer readable memory of claim 14, wherein said preference accumulation module includes executable instructions to deliver selective content to individuals, and instructions to record responses of said individuals to said selective content to form said preference data.

18. The computer readable memory of claim 14, wherein said preference accumulation module includes executable instructions to deliver a Hypertext Markup Language (HTML) page to individuals.

19. The computer readable memory of claim 18, wherein said HTML page includes an image map forming a fine-grained rating menu.

20. The computer readable memory of claim 14, wherein said preference accumulation module includes executable instructions to accumulate a dense matrix of preference data for said set of individuals.

21. The computer readable memory of claim 14, wherein said preference accumulation module includes executable instructions to select predictor set objects with high variance in preference values.