Intelligent system and methods of recommending media content items based on user preferences

• US 20020199186A1
• Filed: 06/21/2002
• Published: 12/26/2002
• Est. Priority Date: 12/21/1999
• Status: Active Grant

• Current Assignee: TiVo Solutions Inc. (Adeia Inc.)
• Original Assignee: TiVo Inc. (Adeia Inc.)
• Inventors: Ali, Kamal, Van Stam, Wijnand
• Granted Patent: US 7,840,986 B2
• Time in Patent Office: Days
• US Class Current: 725/28
• CPC Class Codes: G06F 16/735 Filtering based on additional data, e.g. user or group profiles
• CPC Class Codes: G06F 16/7867 using information manually generated, e.g. tags, keywords, comments, title and artist information, manually generated time, location and usage information, user ratings
• CPC Class Codes: G11B 2220/20 Disc-shaped record carriers
• CPC Class Codes: G11B 27/105 of operating discs
• CPC Class Codes: G11B 27/34 Indicating arrangements indicating means incorporated in magazine or cassette G11B23/046 and G11B23/0875; indicating measured values in general G01D
• CPC Class Codes: H04N 21/252 Processing of multiple end-users' preferences to derive collaborative data
• CPC Class Codes: H04N 21/4147 PVR [Personal Video Recorder] H04N5/76 takes precedence ; arrangements for broadcast with accumulation-type receivers H04H20/40
• CPC Class Codes: H04N 21/4316 for displaying supplemental content in a region of the screen, e.g. an advertisement in a separate window
• CPC Class Codes: H04N 21/4334 Recording operations recording of a television signal H04N5/76; arrangements for recording or accumulating broadcast information or broadcast-related information H04H60/27
• CPC Class Codes: H04N 21/44222 Analytics of user selections, e.g. selection of programs or purchase activity monitoring of user selections in data processing systems G06F11/34; arrangements for monitoring the user's behaviour or opinions in broadcast systems H04H60/33
• CPC Class Codes: H04N 21/443 OS processes, e.g. booting an STB, implementing a Java virtual machine in an STB or power management in an STB
• CPC Class Codes: H04N 21/4532 involving end-user characteristics, e.g. viewer profile, preferences monitoring of user activities for profile generation for accessing a video database G06F16/739; user profiles in network data switching protocols H04L67/306; processing of user preferences or user profiles in wireless networks H04W8/18
• CPC Class Codes: H04N 21/454 Content or additional data filtering, e.g. blocking advertisements
• CPC Class Codes: H04N 21/466 Learning process for intelligent management, e.g. learning user preferences for recommending movies services using the results of monitoring in broadcast systems H04H60/61
• CPC Class Codes: H04N 21/4663 involving probabilistic networks, e.g. Bayesian networks
• CPC Class Codes: H04N 21/47 End-user applications
• CPC Class Codes: H04N 21/47214 for content reservation or setting reminders; for requesting event notification, e.g. of sport results or stock market stock exchange data over packet-switching network H04L12/1804; push services over packet-switching network H04L12/1859; notification of incoming messages in packet switching networks H04L51/224
• CPC Class Codes: H04N 21/4756 for rating content, e.g. scoring a recommended movie
• CPC Class Codes: H04N 21/478 Supplemental services, e.g. displaying phone caller identification, shopping application
• CPC Class Codes: H04N 21/4826 using recommendation lists, e.g. of programs or channels sorted out according to their score
• CPC Class Codes: H04N 21/4828 : for searching program descriptors retrieval of video data G06F16/739
• CPC Class Codes: H04N 21/485 : End-user interface for client configuration
• CPC Class Codes: H04N 21/6543 : for forcing some client operations, e.g. recording remote booting in general G06F9/4416
• CPC Class Codes: H04N 21/812 : involving advertisement data advertising per se G06Q30/02
• CPC Class Codes: H04N 21/8166 : involving executable data, e.g. software arrangements for executing specific programs G06F9/44; broadcasting computer programmes in broadcast systems H04H20/91; involving the movement of software or configuration parameters H04L67/34
• CPC Class Codes: H04N 5/76 : Television signal recording
• CPC Class Codes: H04N 5/782 : on tape
• CPC Class Codes: H04N 7/173 : with two-way working, e.g. subscriber sending a programme selection signal
• CPC Class Codes: H04N 7/17318 : Direct or substantially direct transmission and handling of requests
• CPC Class Codes: H04N 7/17354 : in an intermediate station common to a plurality of user terminals

Abstract

A system and method for making program recommendations to users of a network-based video recording system utilizes expressed preferences as inputs to collaborative filtering and Bayesian predictive algorithms to rate television programs using a graphical rating system. The predictive algorithms are adaptive, improving in accuracy as more programs are rated.

Claims

1. A distributed system for predicting items likely to appeal to a user, based on a client-server collaborative filtering engine, wherein ratings are predicted for said items on the client side using correlation factors downloaded from the server, said correlation factors being computed on the server side from preference profiles anonymously posted to said server from a plurality of clients, said system comprising:

a plurality of clients;

a server side, said clients in intermittent communication with said server side over a network connection;

a list of rated items from each client, wherein said lists are periodically transmitted to said server and aggregated into a single list;

means for filtering said rated items based on frequency;

a matrix for each unique pair of items, wherein said matrix tallies ratings for each item of said pair;

means for computing a correlation between items of said pair from said matrix;

means for filtering non-significant correlations;

a list of correlating items, said list comprising a list of all significant correlations, wherein said list is periodically transmitted to at least one client from said server side; and

client-side means for predicting a rating for an unrated item based on the correlations provided in the list of correlating items.

2. The system of claim 1, wherein said connection between said client and said server is stateless, so that transmission of said rated items list is anonymous.

3. The system of claim 2, wherein said preference profile comprises said list of rated items from said client, and wherein said list of rated items includes each item rated by said user and the corresponding rating assigned by said user.

4. The system of claim 3, wherein said user rates said items according to a scalar system of discrete ratings.

5. The system of claim 4, wherein said means for means for filtering rated items comprises a frequency filter, and wherein said frequency filter filters said rated items in two steps.

6. The system of claim 5, wherein said steps comprise;

monitoring frequency of rated items, and discarding items and corresponding ratings that fall below a threshold frequency; and

monitoring frequency of unique pairs of items, and discarding pairs and corresponding ratings that fall below a threshold frequency.

7. The system of claim 4, wherein dimensions of said pair matrices are n×

n, where n equals the number of discrete ratings.

8. The system of claim 7, wherein said ratings are selected from the group −

3, −

2, −

1, 0, 1, 2, and 3, and wherein a negative value indicates an unfavorable rating, a positive value indicates a favorable rating, and a ‘

0’

indicates a neutral rating.

9. The system of claim 7, wherein n=7.

10. The system of claim 7, wherein pairs of ratings are saved to said pair matrices by:

getting a pair of ratings for a pair of items;

either finding the matrix for said pair, or creating the matrix if it doesn'"'"'t exist;

using the rating pair as an index to the matrix, locating a cell for the rating pair; and

incrementing the cell'"'"'s value by 1.

11. The system of claim 10, wherein a correlation factor is calculated for each matrix.

12. The system of claim 11, where said correlation factor is from the range of approximately −

1 to 1.

13. The system of claim 12, wherein a positive correlation factor indicates a positive correlation and a negative correlation factor indicates an anti-correlation.

14. The system of claim 4, wherein said means for filtering non-significant correlations comprises a correlation filter that filters item pairs that correlate weakly.

15. The system of claim 14, wherein a correlation factor having a non-zero value that approaches zero indicates a weak correlation between items of a pair.

16. The system of claim 4, wherein said client-side means for predicting a rating for an unrated item comprises a prediction engine, wherein said prediction engine predicts a rating for an unrated item by:

getting an unrated item;

searching for the item in the correlating items table;

creating a work list of correlation factors for all items in the table correlating to the unrated item, that have been rated by the user;

insuring that the work list is of a predetermined, fixed length by removing the most weakly correlating items if the list exceeds the predetermined fixed length and padding it with correlation factors of 1 and neutral ratings if the list is shorter than the predetermined fixed length;

multiplying rating*correlation for each item in the work list and summing the products;

dividing the sum of the products by the sum of the absolute values of correlation factors for each item in the work list, wherein a resulting value constitutes a predicted rating for said unrated item.

17. The system of claim 1, wherein said server side comprises a single computer having multiple functional units.

18. The system of claim 1, wherein said server side comprises a single server having a plurality of functional units, wherein separate stages of said system are implemented in separate functional units.

19. The system of claim 1, wherein said server side comprises a plurality of computers and wherein separate stages of said system are implemented on separate computers, said computers being in communication with each other.

20. The system of claim 1, wherein said items comprise items of media content.

21. The system of claim 20, wherein said media content comprises any of network television programming, cable television programming, films, pay-per-view television programming and video.

22. The system of claim 1, said system being a component of a network-based video recording system.

24. The method of claim 23, wherein said preference profile comprises said list of rated items from said client, and wherein said list of rated items includes each item rated by said user and the corresponding rating assigned by said user.

25. The method of claim 24, wherein said user rates said items according to a scalar system of discrete ratings.

26. The method of claim 25, wherein step of filtering said rated items according to frequency comprises the steps of:

providing a frequency filter;

using said frequency filter, monitoring frequency of rated items, and discarding items and corresponding ratings that fall below a threshold frequency; and

monitoring frequency of unique pairs of items, and discarding pairs and corresponding ratings that fall below a threshold frequency.

27. The method of claim 25, wherein dimensions of said pair matrices are n×

n, where n equals the number of discrete ratings.

28. The method of claim 27, wherein n=7.

29. The method of claim 25, wherein said step of providing a matrix for each unique pair of items comprises the steps of:

getting a pair of ratings for a pair of items;

either finding the matrix for said pair, or creating the matrix if it doesn'"'"'t exist;

using the rating pair an in dices to the to the matrix, locating the cell for the rating pair; and

incrementing the cell'"'"'s value by 1.

30. The method of claim 29, wherein a correlation factor is calculated for each matrix.

31. The method of claim 30, wherein said correlation factor is from the range of approximately −

1 to 1.

32. The method of claim 31, wherein a positive correlation factor indicates a positive correlation and a negative correlation factor constitutes an anti-correlation.

33. The method of claim 25, wherein said step of filtering out non-significant correlations comprises the steps of:

providing a correlation filter;

using said correlation filter, filtering out correlations wherein the correlation factor is a non-zero value that approaches zero.

34. The method of claim 25, wherein said step of predicting a rating for an unrated item on the client side comprises the steps of:

providing a client-side prediction engine;

searching for the items in the correlating items table;

creating a work list of correlation factors fro all items in the table correlating to the unrated item, that have been rated by the user;

insuring that the work list is of a predetermined, fixed length;

multiplying rating*correlation for each item in the work list, and summing the products; and

dividing the sum of the products by the sum of the absolute values of correlation factors for each item in the work list, wherein a resulting value constitutes a predicted rating for said unrated item.

35. The method of claim 34, wherein said step of insuring that the work list is of a predetermined, fixed length comprises one of the steps of:

removing the most weakly correlating items if the list exceeds the predetermined fixed length; and

padding the list with correlation factors of one and neutral ratings if the list is shorted than the predetermined, fixed length.

36. The method of claim 23, wherein said server side comprises a single server having a plurality of functional units, wherein separate stages of said method are executed in separate functional units.

37. The method of claim 23, wherein said server side comprises a plurality of computers and wherein separate stages of said method are executed on separate computers, said computers being in communication with each other.

38. The method of claim 23, wherein said items comprise items of media content;

39. The method of claim 38, wherein said media content comprises any of network television programming, cable television programming, films, pay-per-view television programming, and video.

40. The method of claim 23, wherein said method is implemented within a network-based video recording system.

23. A method of predicting items likely to appeal to a user, based on a distributed collaborative filtering engine, wherein ratings are predicted for said items on the client side using correlation factors downloaded from the server, said correlation factors being computed on the server side from preference profiles anonymously posted to said server from a plurality of clients, said method comprising the steps of:

providing a server;

providing a plurality of clients, wherein said clients are intermittently in contact with said server;

providing a population of items, wherein said user has rated a portion of said population;

periodically transmitting a list of said rated items to said server side over a stateless network connection;

aggregating said list with lists from other clients into a single list;

filtering said list based on frequency;

providing a matrix for each unique pair of items, wherein said matrix tallies ratings for each item of said pair;

computing a correlation from said matrix between items of said pair;

filtering out non-significant correlations;

compiling a list of correlating items comprising a list of all non-significant correlations;

periodically transmitting said list to said clients;

on the client side, predicting a rating for an unrated item, based on the correlation provided in the list of correlating items.