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

• US 20020199194A1
• Filed: 06/14/2002
• 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
• Granted Patent: US 8,079,046 B2
• Time in Patent Office: Days
• US Class Current: 725/46
• 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 method of predicting items of media content most likely to appeal to a user and presenting said predictions in a graphical user interface wherein said predictions are based on preferences expressed by said user, comprising the steps of:

providing a rating system;

providing a metaphor for expressing said rating system;

providing an interactive user interface for assigning ratings by said user;

rating items according to said rating system by said user using said interactive interface;

generating ratings of new content items according to one or more predictive algorithms, based on said user ratings;

displaying said user ratings and said generated ratings in said graphical user interface.

2. The method of claim 1, wherein items are defined by predetermined attributes, each attribute having a plurality of possible values, and wherein said step of providing an interactive user interface for assigning ratings by said user comprises the steps of:

providing interactive lists of items; and

providing interactive lists of attributes and corresponding values, wherein each unique attribute-value pair constitutes a feature.

3. The method of claim 2, wherein said step of rating items according to said rating system by said user comprises at least one of the steps of

assigning an overall rating to an item;

assigning ratings to individual features.

4. The method of claim 3, wherein said rating system includes values from a range of approximately −

3 to 3, so that a negative number signifies an unfavorable rating, a positive number signifies a favorable rating and a ‘

0’

signifies a neutral rating.

5. The method of claim 4, wherein said graphical metaphor utilizes “

thumbs up”

icons for favorable ratings and “

thumbs down”

icons for unfavorable ratings, and wherein the number of thumbs assigned to an item equals the numerical value of said rating.

6. The method of claim 5, wherein said step of generating ratings comprises the steps of:

providing a collaborative filtering algorithm;

identifying new items that correlate highly with items previously rated by said user, according to said collaborative filtering algorithm; and

assigning ratings to said new items based on said correlation and said previous user rating.

7. The method of claim 5, wherein said step of generating ratings comprises the steps of:

providing a content-based adaptive modeling algorithm;

evaluating user-assigned ratings of individual features;

assigning an overall rating to a new item defined by any of said user-rated features based on said user-assigned ratings.

8. The method of claim 7, wherein said step of generating ratings comprises the steps of:

providing a modified naï

ve Bayes classifier;

pooling all features defining items assigned overall ratings by said user, except for featrues that have explicitly been rated by said user;

tallying the occurrence of each distinct feature in said pool, and weighting it according to the overall ratings of the items it defines, according to said naï

ve Bayes classifier; and

generating overall ratings for new items based on said tallies and weightings of said features.

9. The method of claim 5, wherein said media content comprises any of network television programming, cable television programming, network television programming, pay-per-view television programming, movies and public-access television programming.

10. The method of claim 9, wherein said attributes include:

genre, actor and director.

11. The method of claim 5, wherein the “

thumbs up” and

“

thumbs down”

icons for user assigned ratings are visually distinct from those for generated ratings.

12. The method of claim 5, wherein confidence level of said predictive algorithms increases as the number of explicit ratings increases.

13. The method of claim 12, wherein the step of displaying said ratings comprises the steps of:

aggregating said user-rated items and outputs of said predictive algorithms into a single list;

sorting said list according to rating in descending order;

sorting items having the same rating according to confidence level in descending order; and

displaying said sorted list in said graphical user interface as a list, wherein an item is displayed accompanied by its rating icon.

14. The method of claim 12, further comprising the step of

changing a rating by said user;

and

updating the rating icon to correspond to said changed rating.

15. The method of claim 14, wherein a user-assigned rating cannot be replaced by a generated rating.

17. The method of claim 16, wherein said collaborative filtering algorithm rates items according to correlation factors, so that an item with a high correlation factor to an item rated by said user is similarly rated.

18. The method of claim 17, wherein said correlation factors are provided in a table of correlating items.

19. The method of claim 18, wherein said table of correlating items comprises a plurality of pairs of items and a corresponding correlation factor for each pair.

20. The method of claim 17, wherein said correlation factors are calculated from other users'"'"' ratings, wherein said other users'"'"' ratings are stored in one or more pair matrices, and wherein each matrix yields a correlation factor in a range of approximately −

1 to 1.

21. The method of claim 18, wherein said table of correlating items is downloaded to a client from a server.

22. The method of claim 16, wherein said step of rating a new item according to a second algorithm comprises the steps of:

presenting said user with a list of possible features;

assigning ratings to at least a portion of the displayed features by said user;

inputting said ratings of said values to said second algorithm, wherein said algorithm rates new items, wherein features high specificity that occur infrequently are heavily weighted and considered to be reliable predictors.

23. The method of claim 22, wherein said step of presenting said user with a list of possible features comprises:

providing an interactive user interface, wherein each attribute is displayed as a user selection on a screen, wherein selecting an attribute navigates the user to a subsequent screen displaying one or more lists of values, each unique attribute-value pair constituting a feature, wherein each list is a user selection.

24. The method of claim 23, wherein said step of assigning ratings to at least a portion of said values by said user comprises the steps of:

selecting a list of features, wherein features constituting said list are displayed on a subsequent screen; and

assigning ratings to at least a portion of said displayed features.

25. The method of claim 16, wherein said step of rating new items based on previous explicit user ratings of items comprises the steps of:

determining features that correlate to said user'"'"'s tastes and preferences, based on previous ratings of items by said user;

optionally, presenting said user with a list of said features so that said user may interactively modify said list; and

inputting said features to said second algorithm, wherein said algorithm rates items not previously rated.

26. The method of claim 25, wherein said second algorithm includes a modified naï

ve Bayes classifier, wherein said modified Bayes classifier accommodates multiple simultaneous features, and wherein said modified Bayes classifier is capable of excluding selected features.

27. The method of claim 26, wherein user ratings comprise explicit ratings and generated ratings comprise implicit ratings, and wherein an explicit rating overrides an implicit rating.

28. The method of claim 27, wherein explicitly rated features are excluded as inputs to said Bayes classifier so that said algorithm works only with unrated features, said unrated features comprising free-floating features.

29. The method of claim 28, wherein said step of determining features comprises the steps of:

aggregating all free-floating features from said user-rated items into a population of features;

tallying the occurrences of unique features in said population;

assigning a weight to each unique feature, wherein a featureof high-specificity has a higher weight than one of low specificity; and

wherein an infrequently-occurring, highly-weighted feature constitutes a reliable predictor of items likely to appeal to said user.

30. The method of claim 25, wherein said step of presenting said user with a list of said attribute-value pairs so that said user may interactively modify said list comprises providing an interactive user interface, wherein each attribute is displayed as a user selection on a screen, wherein selecting an attribute navigates the user to a subsequent screen displaying one or more lists of values, wherein each list is a user selection.

31. The method of claim 30, wherein said step of presenting said user with a list of said attribute-value pairs so that said user may interactively modify said list further comprises the steps of:

selecting a list of features, wherein features constituting said list are displayed on a subsequent screen;

evaluating ratings assigned to said features; and

optionally, modifying said ratings.

32. The method of claim 16, wherein said media content comprises any of network television programming, cable television programming, network television programming, pay-per-view television programming, movies and public-access television programming.

33. The method of claim 16, wherein said attributes include:

director, actor and genre.

34. The method of claim 16, wherein said ratings include values from a range of approximately −

3 to 3, so that a negative number signifies an unfavorable rating, a positive number signifies a favorable rating and a ‘

0’

signifies a neutral rating.

35. The method of claim 34, wherein a graphical metaphor expresses said ratings, so that a “

thumbs up”

icon signifies a favorable rating and a “

thumbs down”

icon signifies an unfavorable rating, and wherein the number of thumbs assigned is equal to the absolute value of said rating.

36. The method of claim 35, wherein icons signifying explicit ratings are distinct from icons signifying implicit ratings.

37. The method of claim 36, wherein confidence level of said predictive algorithms increases as the number of explicit user ratings increases.

39. The method of claim 38, wherein said step of predicting a rating based on features explicitly rated by said user comprises the steps of:

inputting ratings of features defining said item to said algorithm;

rating said item according to said features, wherein features of high specificity are heavily weighted, and wherein an infrequently-occurring feature of high specificity constitutes a reliable predictor for said item.

40. The method of claim 38, wherein said step of predicting a rating based on implicitly rated features comprises the steps of:

determining features that correlate to said user'"'"'s tastes and preferences, based on previous ratings of items by said user;

optionally, presenting said user with a list of said features so that said user may interactively modify said list; and

inputting said features to said adaptive modeling algorithm, wherein said algorithm rates items not previously rated.

41. The method of claim 40, wherein said adaptive modeling algorithm includes a modified naï

ve Bayes classifier, wherein said modified Bayes classifier accommodates multiple simultaneous features, and wherein said modified Bayes calssifier is capable of excluding selected features.

42. The method of claim 39, wherein an explicit rating overrides an implicit rating.

43. The method of claim 42, wherein explicitly rated features are excluded as inputs to said Bayes classifier so that said algorithm works only with unrated features, said unrated features constituting free-floating features.

44. The method of claim 43, wherein said step of determining features comprises the steps of:

aggregating all free-floating features from said user-rated items into a population of features;

tallying the occurrences of unique features in said population;

assigning a weight to each unique feature, wherein a feature of high-specificity has a higher weight than one of low specificity; and

wherein an infrequently-occurring, highly-weighted feature constitutes a reliable predictor of items likely to appeal to said user.

45. The method of claim 38, wherein said step of presenting said user with a list of said attribute-value pairs so that said user may interactively modify said list comprises providing an interactive user interface, wherein each attribute is displayed as a user selection on a screen, wherein selecting an attribute navigates the user to a subsequent screen displaying one or more lists of values, wherein each list is a user selection.

46. The method of claim 45, wherein said step of presenting said user with a list of said features so that said user may interactively modify said list further comprises the steps of:

selecting a list of features, wherein features constituting said list are displayed on a subsequent screen;

evaluating ratings assigned to said features; and

optionally, modifying said ratings.

47. The method of claim 38, wherein said media content comprises any of network television programming, cable television programming, network television programming, pay-per-view television programming, movies and public-access television programming.

48. The method of claim 38, wherein said ratings include values from a range of approximately −

3 to 3, so that a negative number signifies an unfavorable rating, a positive number signifies a favorable rating and a ‘

0’

signifies a neutral rating.

49. The method of claim 48, wherein confidence level of said adaptive modeling algorithm increases with increasing numbers of explicitly rated items.

50. The method of claim 49, wherein a graphical metaphor expresses said ratings, so that a “

thumbs up”

icon signifies a favorable rating and a “

thumbs down”

icon signifies an unfavorable rating, and wherein the number of thumbs assigned is equal to the absolute value of said rating.

51. The method of claim 50, wherein icons signifying explicit ratings are distinct from icons signifying implicit ratings.

52. The method of claim 50, further comprising the steps of:

aggregating user-rated items and output of said algorithm into a single list;

sorting said list according to rating in descending order;

sorting items having the same rating according to confidence level in descending order; and

displaying said list to said in a graphical user interface.

16. A method of predicting items of media content likely to appeal to a user, based on said user'"'"'s preferences, comprising at least one of the steps of:

rating a new item based on earlier user ratings of items that correlate highly with said new item according to a first predictive algorithm, said first algorithm comprising a collaborative filtering algorithm;

rating a new item according to a second predictive algorithm, said second algorithm comprising a content-based adaptive modeling algorithm, based on user ratings of specific features describing said item, wherein at least a portion of said features have previously been explicitly rated by said user; and

for a new item whose features have not been previously rated by said user, rating said new item based on previous explicit user ratings of items, according to said adaptive modeling algorithm, wherein said algorithm generates implicit ratings of features based on said explicit user ratings.

38. The method of claim 37, further comprising the steps of:

aggregating user-rated items and output from said predictive algorithms into a single list;

sorting said list according to rating in descending order;

sorting items having the same rating according to confidence level in descending order; and

displaying said list to said user as a list in a graphical user interface.

38-1. A method of predicting items of media content likely to appeal to a user, wherein a new item is rated based on ratings of features describing said item according to a content-based adaptive modeling algorithm, said method comprising one of the steps of:

predicting a rating based on features explicitly rated by said user; and

in the absence of explicitly rated features, predicting a rating based on implicitly rated features, wherein said adaptive modeling algorithm generates said implicit ratings based on previous overall ratings of items by said user.