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

• US 8,079,046 B2
• Filed: 12/14/2000
• Issued: 12/13/2011
• Est. Priority Date: 12/21/1999
• Status: Expired due to Fees

• Current Assignee: TiVo Solutions Inc. (Adeia Inc.)
• Original Assignee: TiVo Inc. (Adeia Inc.)
• Inventors: Ali, Kamal, Van Stam, Wijnand
• Primary Examiner(s): Kincaid, Kristine
• Assistant Examiner(s): Parra, Omar S
• Application Number:
  US10/168,070
• Publication Number:
  US 20020199194A1
• Time in Patent Office: 4,016 Days
• Field of Search: 725/46, 725/47, 725/58, 725/9, 725/13, 725/24, 725/28, 725/87, 725/114, 725/131
• 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, comprising:

accepting ratings, at a multimedia recording device, of media content items from a user, via a graphical user interface, according to a rating system;

expressing the rating system using a graphical metaphor;

receiving by the multimedia recording device, from a server, correlation data between pairs of available media content;

generating ratings at the multimedia recording device of media content items not rated by the user according to one or more predictive algorithms, based on the media content items specifically rated by the user and the correlation data received from the server;

displaying, by the multimedia recording device, the media content items ratings by the user and the generated media content ratings in the graphical user interface; and

allowing the user to correct generated media content ratings displayed in the graphical user interface in order to obtain generated media content ratings more in line with what the user expects.

2. The method of claim 1, wherein items are defined by predetermined attributes, each attribute having a plurality of possible values, and wherein the rating step further comprises:

displaying interactive lists of items; and

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

3. The method of claim 2, wherein the attributes include any of:

genre, actor, and director.

4. The method of claim 1, wherein the step of rating items according to the rating system by the user comprises at least one of the steps of:

assigning an overall rating to an item; and

assigning ratings to individual features.

5. The method of claim 1, wherein the rating system includes values from a range of approximately −

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

0’

signifies a neutral rating.

6. The method of claim 1, wherein the graphical metaphor utilises “

thumbs up”

icons for favorable ratings and “

thumbs down”

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

7. The method of claim 6, wherein the “

thumbs up” and

“

thumbs down”

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

8. The method of claim 6, further comprising:

changing a rating by the user; and

updating a rating icon to correspond to the changed rating.

9. The method of claim 8, wherein a user-assigned rating cannot be replaced by a generated rating.

10. The method of claim 1, wherein the step of generating ratings comprises:

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

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

11. The method of claim 1, wherein the step of generating ratings comprises:

implementing a content-based adaptive modeling algorithm;

evaluating user-assigned ratings of individual features; and

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

12. The method of claim 11, wherein the step of generating ratings comprises:

implementing a modified naï

ve Bayes classifier;

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

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

ve Bayes classifier; and

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

13. The method of claim 1, wherein the media content comprises any of:

network television programming, cable television programming, pay-per-view television programming, movies and public-access television programming.

14. The method of claim 1, wherein confidence level of the predictive algorithms increases as a number of explicit ratings increases.

15. The method of claim 1, wherein the step of displaying the ratings comprises:

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

sorting the list according to rating in descending order;

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

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

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

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

rating the new item, when the new item has not been rated by the first predictive algorithm, according to a second predictive algorithm, the second algorithm comprising a content-based adaptive modeling algorithm, based on ratings of specific features by the user describing the item, wherein at least a portion of the features have previously been explicitly rated by the user;

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

allowing the user to correct a rating for an item generated by the first predictive algorithm or the second predictive algorithm in order to obtain generated item ratings more in line with what the user expects;

wherein the method is performed by one or more computing devices.

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

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

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

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

21. The method of claim 17, wherein the correlation factors are calculated from other users'"'"' ratings, wherein the 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.

22. The method of claim 16, further comprising:

presenting the user with a list of possible features;

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

inputting the ratings of the values to the second predictive algorithm, wherein features of high-specificity that occur infrequently are heavily weighted and considered to be reliable predictors.

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

wherein descriptive attributes are displayed as user selections 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.

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

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

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

25. The method of claim 23, wherein the attributes include any of:

director, actor and genre.

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

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

optionally, presenting the user with a list of said features allowing the user to interactively modify the list; and

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

27. The method of claim 26, wherein the second predictive algorithm includes a modified naï

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

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

29. The method of claim 28, wherein explicitly rated features are excluded as inputs to the Bayes classifier whereby the algorithm works only with unrated features, the unrated features comprising free-floating features.

30. The method of claim 29, wherein the step of determining features comprises:

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

tallying occurrences of unique features in the 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 the user.

31. The method of claim 26, wherein the step of presenting the user with a list of the features allowing the user to interactively modify the list further comprises:

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

allowing the user to modify the ratings.

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

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

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

0’

signifies a neutral rating.

34. The method of claim 33, wherein a graphical metaphor expresses the ratings, wherein a “

thumbs up”

icon signifies a favorable rating and a “

thumbs down”

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

35. The method of claim 34, wherein icons signifying explicit ratings are distinct from icons signifying implicit ratings.

36. The method of claim 16, wherein confidence level of the predictive algorithms increases as the number of explicit user ratings increases.

37. The method of claim 16, further comprising:

aggregating user-rated items and generated item ratings from the first predictive algorithm and the second predictive algorithm into a single list; and

displaying the list to the user.

38. A method of predicting items of media content likely to appeal to a user, comprising:

predicting a rating of an item based on features explicitly rated by a user;

in the absence of explicitly rated features, predicting a rating of an item based on implicitly rated features, wherein an adaptive modeling algorithm generates the implicit feature ratings based on previous overall ratings of items made by the user; and

allowing the user to correct a predicted rating for an item in order to obtain predicted item ratings more in line with what the user expects;

wherein the method is performed by one or more computing devices.

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

inputting features defining the item to the adaptive modeling algorithm; and

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

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

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

optionally, presenting the user with a list of the features allowing the user to interactively modify the list; and

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

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

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

42. The method of claim 41 wherein explicitly rated features are excluded as inputs to the Bayes classifier whereby the adaptive modeling algorithm works only with unrated features, the unrated features constituting free-floating features.

43. The method of claim 42, wherein the step of determining features comprises:

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

tallying occurrences of unique features in the 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 the user.

44. The method of claim 40, wherein the step of presenting the user with a list of the features allowing the user to interactively modify the list further comprises:

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

allowing the user to modify the ratings.

45. The method of claim 38, wherein an explicit rating overrides an implicit rating.

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

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

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

0’

signifies a neutral rating.

48. The method of claim 47, wherein a graphical metaphor expresses the ratings, wherein 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 the rating.

49. The method of claim 48, wherein icons signifying explicit ratings are distinct from icons signifying implicit ratings.

50. The method of claim 48, further comprising:

aggregating user-rated items and output of the adaptive modeling algorithm into a single list; and

displaying the list to the user.

51. The method of claim 38, wherein confidence level of the adaptive modeling algorithm increases with increasing numbers of explicitly rated items.