Approach for estimating user ratings of items

• US 20050125307A1
• Filed: 02/06/2004
• Published: 06/09/2005
• Est. Priority Date: 04/28/2000
• Status: Active Grant

• Current Assignee: Netflix, Inc.
• Original Assignee: Netflix, Inc.
• Inventors: Shih, Lawrence Wen-Kai, Lanning, Stanley Miguel, Purcell, Shawn Michael, Hunt, Neil Duncan, Ciancutti, John Robert, Hastings, W. Reed
• Granted Patent: US 7,617,127 B2
• Time in Patent Office: Days
• US Class Current: 705/26
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data
• CPC Class Codes: G06Q 30/0204 Market segmentation
• CPC Class Codes: G06Q 30/0217 involving input on products or services in exchange for incentives or rewards
• CPC Class Codes: G06Q 30/06 Buying, selling or leasing transactions
• CPC Class Codes: G06Q 30/0623 Item investigation
• CPC Class Codes: G06Q 40/04 Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange

Abstract

According to another embodiment of the invention, an approach is provided for estimating how a particular user would rate a particular item from a plurality of items. The approach is applicable to any type of items, including rental items such as movies, music and games, and the invention is not limited to any particular type of item. One or more items from the plurality of items that have ratings similar to the particular item are identified. Then, one or more other users are identified that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items. Finally, an estimation is made of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.

Claims

1. A machine-implemented method for estimating how a particular user would rate a particular item from a plurality of items, the method comprising the machine-implemented steps of:

identifying, from the plurality of items, one or more other items that have ratings similar to the particular item;

selecting a set of other users from a plurality of other users based upon user selection criteria;

identifying one or more other users from the set of other users that have given ratings to a subset of the one or more other items, wherein the ratings are statistically similar to ratings given by the particular user to the subset of the one or more other items; and

generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.

2. The machine implemented method as recited in claim 1, wherein the step of identifying, from the plurality of items, one or more other items that have ratings similar to the particular item is performed off-line in advance of the other steps.

3. A machine-readable medium for estimating how a particular user would rate a particular item from a plurality of items, the machine-readable medium carrying instructions which, when processed by one or more processors, cause performance of the steps of:

identifying, from the plurality of items, one or more other items that have ratings similar to the particular item;

selecting a set of other users from a plurality of other users based upon user selection criteria;

identifying one or more other users from the set of other users that have given ratings to a subset of the one or more other items, wherein the ratings are statistically similar to ratings given by the particular user to the subset of the one or more other items; and

generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.

4. The machine-readable medium as recited in claim 3, wherein the step of identifying, from the plurality of items, one or more other items that have ratings similar to the particular item is performed off-line in advance of the other steps.

5. An apparatus for estimating how a particular user would rate a particular item from a plurality of items, the apparatus comprising a memory storing instructions which, when processed by one or more processors, cause performance of the steps of:

identifying, from the plurality of items, one or more other items that have ratings similar to the particular item;

selecting a set of other users from a plurality of other users based upon user selection criteria;

identifying one or more other users from the set of other users that have given ratings to a subset of the one or more other items, wherein the ratings are statistically similar to ratings given by the particular user to the subset of the one or more other items; and

generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.

6. The apparatus as recited in claim 5, wherein the step of identifying, from the plurality of items, one or more other items that have ratings similar to the particular item is performed off-line in advance of the other steps.

7. A machine-implemented method for estimating how a particular user would rate a particular item from a plurality of items, the method comprising the machine-implemented steps of:

identifying one or more items from the plurality of items that have ratings similar to the particular item;

identifying one or more other users that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items; and

generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.

8. The machine implemented method as recited in claim 7, wherein the step of identifying, from the plurality of items, one or more other items that have ratings similar to the particular item is performed off-line in advance of the other steps.

9. The machine-implemented method as recited in claim 7, wherein identifying one or more items from a plurality of items that have ratings similar to the particular item includes calculating a statistical correlation between the particular item and the other items from the plurality of items.

10. The machine-implemented method as recited in claim 7, further comprising the machine-implemented step of determining whether the particular item and each of the one or more items have at least a specified number of co-ratings.

11. The machine-implemented method as recited in claim 7, further comprising the machine-implemented step of determining whether the particular item and one of the one or more items have at least a specified fraction of co-ratings between the particular item and the one of the one or more items divided by the number of ratings of either the particular item or the one of the one or more items.

12. The machine-implemented method as recited in claim 7, wherein the step of identifying one or more items from the plurality of items that have ratings similar to the particular item includes identifying whether the particular item and the one or more items have a number of co-raters greater than a specified fraction of a statistically expected number, given the number of ratings of each item counted separately.

13. The machine-implemented method as recited in claim 7, wherein the step of identifying one or more items from the plurality of items is performed by one or more external agents.

14. The machine-implemented method as recited in claim 7, wherein the step of identifying one or more items from the plurality of items that have ratings similar to the particular item includes consideration of one or more of a statistical correlation between the particular item and the one or more items, whether the particular item and the one or more items have at least a specified number of co-ratings, whether the particular item and the one or more items have at least a specified fraction of co-ratings, and whether the particular item and the one or more items have at least a specified fraction of the statistically expected number of co-ratings.

15. The machine-implemented method as recited in claim 7, further comprising the machine-implemented step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having at least a specified minimum number of ratings.

16. The machine-implemented method-as recited in claim 7, further comprising the machine-implemented step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having less than a specified maximum number of ratings.

17. The machine-implemented method as recited in claim 7, further comprising the machine-implemented step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having an average rating within a specified range.

18. The machine-implemented method as recited in claim 7, further comprising the machine-implemented step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having a rating distribution within a specified tolerance of a Normal distribution.

19. The machine-implemented method as recited in claim 7, further comprising the machine-implemented step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having a rating standard deviation within a specified range.

20. The machine-implemented method as recited in claim 7, wherein identifying one or more other users that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items includes consideration of one or more of whether the one or more other users have at least a specified minimum number of co-ratings, whether the one or more other users have less than a specified maximum number of ratings, whether the one or more other users have an average rating within a specified range, whether the one or more other users have a rating distribution within a specified tolerance of a Normal distribution, and whether the one or more other users have a rating standard deviation within a specified range.

21. The machine-implemented method as recited in claim 7, wherein identifying the one or more other users includes choosing the one or more other users for whom a statistical correlation with the particular user of co-ratings of the one or more items is greater than a specified threshold.

22. The machine-implemented method as recited in claim 7, wherein identifying the one or more other users includes choosing the one or more other users for whom a vector cosine distance of co-ratings of the one or more items is smaller than a specified threshold.

23. The machine-implemented method as recited in claim 7, wherein identifying the one or more other users includes choosing the one or more other users who have co-rated a subset of the one or more items exactly the same as the particular user.

24. The machine-implemented method as recited in claim 7, wherein identifying the one or more other users includes choosing the one or more other users who have rated co-rated a subset of the one or more items within a specified threshold of the ratings by the particular user.

25. The machine-implemented method as recited in claim 7, wherein identifying the one or more other users includes consideration of one or more of whether the one or more users have a correlation of co-ratings of the one or more items greater than a specified threshold, whether the one or more users have a vector cosine distance of co-ratings of the one or more items smaller than a specified threshold, whether the one or more users have co-rated a subset of the one or more items exactly the same, and whether the one or more users having co-rated a subset of the one or more items within a specified threshold of the ratings by the particular user.

26. The machine-implemented method as recited in claim 7, wherein identifying the one or more other users includes choosing a specified number of other users most similar to the particular user.

27. The machine-implemented method as recited in claim 7, wherein identifying the one or more other users includes choosing other users whose similarity to the particular user is better than a specified threshold.

28. The machine-implemented method as recited in claim 7, wherein identifying the one or more other users includes choosing up to a specified number of other users most similar to the particular user, where all the chosen other users have similarity better than a specified threshold.

29. The machine-implemented method as recited in claim 7, wherein generating an estimation of how the user would rate the particular item based upon ratings for the particular item given by the one or more other users includes determining an average rating of the particular item by the one or more other users.

30. The machine-implemented method as recited in claim 7, wherein generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users includes determining an average rating of the particular item by the one or more other users.

31. The machine-implemented method as recited in claim 7, further comprising the machine-implemented steps of:

generating, for each of the one or more users, a separate estimation of how the particular user would rate the particular item based upon the rating for the particular item given by each of the one or more other users, and

calculating a confidence metric based upon disparities between the separate estimations.

32. The machine-implemented method as recited in claim 7, wherein the plurality of items is a plurality of movies, the one or more items is one or more movies and the particular item is a particular movie.

33. The machine-implemented method as recited in claim 7, wherein the plurality of items is a plurality of music items, the one or more items is one or more music items and the particular item is a particular music item.

34. A machine-readable medium for estimating how a particular user would rate a particular item from a plurality of items, the machine-readable medium carrying instructions which, when processed by one or more processors, cause performance of the steps of:

identifying, from the plurality of items, one or more items that have ratings similar to the particular item;

identifying one or more other users that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items; and

generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.

35. The machine-readable medium as recited in claim 34, wherein the step of identifying, from the plurality of items, one or more other items that have ratings similar to the particular item is performed off-line in advance of the other steps.

36. The machine-readable medium as recited in claim 34, wherein identifying one or more items from a plurality of items that have ratings similar to the particular item includes calculating a statistical correlation between the particular item and the other items from the plurality of items.

37. The machine-readable medium as recited in claim 34, further comprising one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of determining whether the particular item and each of the one or more items have at least a specified number of co-ratings.

38. The machine-readable medium as recited in claim 34, further comprising one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of determining whether the particular item and one of the one or more items have at least a specified fraction of co-ratings between the particular item and the one of the one or more items divided by the number of ratings of either the particular item or the one of the one or more items.

39. The machine-readable medium as recited in claim 34, wherein the step of identifying one or more items from the plurality of items that have ratings similar to the particular item includes identifying whether the particular item and the one or more items have a number of co-raters greater than a specified fraction of a statistically expected number, given the number of ratings of each item counted separately.

40. The machine-readable medium as recited in claim 34, wherein the step of identifying one or more items from the plurality of items is performed by one or more external agents.

41. The machine-readable medium as recited in claim 34, wherein the step of identifying one or more items from the plurality of items that have ratings similar to the particular item includes consideration of one or more of a statistical correlation between the particular item and the one or more items, whether the particular item and the one or more items have at least a specified number of co-ratings, whether the particular item and the one or more items have at least a specified fraction of co-ratings, and whether the particular item and the one or more items have at least a specified fraction of the statistically expected number of co-ratings.

42. The machine-readable medium as recited in claim 34, further comprising one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having at least a minimum number of ratings.

43. The machine-readable medium as recited in claim 34, further comprising one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having less than a maximum number of ratings.

44. The machine-readable medium as recited in claim 34, further comprising one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having an average rating within a specified range.

45. The machine-readable medium as recited in claim 34, further comprising one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having a rating distribution within a specified tolerance of a Normal distribution.

46. The machine-readable medium as recited in claim 34, further comprising one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having a rating standard deviation within a specified range.

47. The machine-implemented method as recited in claim 34, wherein identifying one or more other users that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items includes consideration of one or more of whether the one or more other users have at least a specified minimum number of co-ratings, whether the one or more other users have less than a specified maximum number of ratings, whether the one or more other users have an average rating within a specified range, whether the one or more other users have a rating distribution within a specified tolerance of a Normal distribution, and whether the one or more other users have a rating standard deviation within a specified range.

48. The machine-implemented method as recited in claim 34, wherein identifying the one or more other users includes choosing the one or more other users for whom a statistical correlation with the particular user of co-ratings of the one or more items is greater than a specified threshold.

49. The machine-implemented method as recited in claim 34, wherein identifying the one or more other users includes choosing the one or more other users for whom a vector cosine distance of co-ratings of the one or more items is smaller than a specified threshold.

50. The machine-implemented method as recited in claim 34, wherein identifying the one or more other users includes choosing the one or more other users who have co-rated a subset of the one or more items exactly the same as the particular user.

51. The machine-implemented method as recited in claim 34, wherein identifying the one or more other users includes choosing the one or more other users who have rated co-rated a subset of the one or more items within a specified threshold of the ratings by the particular user.

52. The machine-implemented method as recited in claim 34, wherein identifying the one or more other users includes consideration of one or more of whether the one or more users have a correlation of co-ratings of the one or more items greater than a specified threshold, whether the one or more users have a vector cosine distance of co-ratings of the one or more items smaller than a specified threshold, whether the one or more users have co-rated a subset of the one or more items exactly the same, and whether the one or more users having co-rated a subset of the one or more items within a specified threshold of the ratings by the particular user.

53. The machine-implemented method as recited in claim 34, wherein identifying the one or more other users includes choosing a specified number of other users most similar to the particular user.

54. The machine-implemented method as recited in claim 34, wherein identifying the one or more other users includes choosing other users whose similarity to the particular user is better than a specified threshold.

55. The machine-implemented method as recited in claim 34, wherein identifying the one or more other users includes choosing up to a specified number of other users most similar to the particular user, where all the chosen other users have similarity better than a specified threshold.

56. The machine-readable medium as recited in claim 34, wherein generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users includes determining an average rating of the particular item by the one or more other users.

57. The machine-readable medium as recited in claim 34, wherein generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users includes determining an average rating of the particular item by the one or more other users.

58. The machine-readable medium as recited in claim 34, further comprising one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional steps of:

generating, for each of the one or more users, a separate estimation of how the particular user would rate the particular item based upon the rating for the particular item given by each of the one or more other users, and

calculating a confidence metric based upon disparities between the separate estimations.

59. The machine-readable medium as recited in claim 34, wherein the plurality of items is a plurality of movies, the one or more items is one or more movies and the particular item is a particular movie.

60. The machine-readable medium as recited in claim 34, wherein the plurality of items is a plurality of music items, the one or more items is one or more music items and the particular item is a particular music item.

61. An apparatus for estimating how a particular user would rate a particular item from a plurality of items, the apparatus comprising a memory storing instructions which, when processed by one or more processors, cause the one or more processors to perform the steps of:

identifying, from the plurality of items, one or more items that have ratings similar to the particular item;

identifying one or more other users that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items; and

generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.

62. The apparatus as recited in claim 61, wherein the step of identifying, from the plurality of items, one or more other items that have ratings similar to the particular item is performed off-line in advance of the other steps.

63. The apparatus as recited in claim 61, wherein identifying one or more items from a plurality of items that have ratings similar to the particular item includes calculating a statistical correlation between the particular item and the other items from the plurality of items.

64. The apparatus as recited in claim 61, wherein the memory further comprises one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of determining whether the particular item and each of the one or more items have at least a specified number of co-ratings.

65. The apparatus as recited in claim 61, wherein the memory further comprises one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of determining whether the particular item and one of the one or more items have at least a specified fraction of co-ratings between the particular item and the one of the one or more items divided by the number of ratings of either the particular item or the one of the one or more items.

66. The apparatus as recited in claim 61, wherein the step of identifying one or more items from the plurality of items that have ratings similar to the particular item includes identifying whether the particular item and the one or more items have a number of co-raters greater than a specified fraction of a statistically expected number, given the number of ratings of each item counted separately.

67. The apparatus as recited in claim 61, wherein the step of identifying one or more items from the plurality of items is performed by one or more external agents.

68. The apparatus as recited in claim 61, wherein the step of identifying one or more items from the plurality of items that have ratings similar to the particular item includes consideration of one or more of a statistical correlation between the particular item and the one or more items, whether the particular item and the one or more items have at least a specified number of co-ratings, whether the particular item and the one or more items have at least a specified fraction of co-ratings, and whether the particular item and the one or more items have at least a specified fraction of the statistically expected number of co-ratings.

69. The apparatus as recited in claim 61, wherein the memory further comprises one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having at least a minimum number of ratings.

70. The apparatus as recited in claim 61, wherein the memory further comprises one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having less than a maximum number of ratings.

71. The apparatus as recited in claim 61, wherein the memory further comprises one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having an average rating within a specified range.

72. The apparatus as recited in claim 61, wherein the memory further comprises one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having a rating distribution within a specified tolerance of a Normal distribution.

73. The apparatus as recited in claim 61, wherein the memory further comprises one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional step of selecting the one or more other users from a plurality of other users based upon each of the one or more other users having a rating standard deviation within a specified range.

74. The apparatus as recited in claim 61, wherein identifying one or more other users that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items includes consideration of one or more of whether the one or more other users have at least a specified minimum number of co-ratings, whether the one or more other users have less than a specified maximum number of ratings, whether the one or more other users have an average rating within a specified range, whether the one or more other users have a rating distribution within a specified tolerance of a Normal distribution, and whether the one or more other users have a rating standard deviation within a specified range.

75. The apparatus as recited in claim 61, wherein identifying the one or more other users includes choosing the one or more other users for whom a statistical correlation with the particular user of co-ratings of the one or more items is greater than a specified threshold.

76. The apparatus as recited in claim 61, wherein identifying the one or more other users includes choosing the one or more other users for whom a vector cosine distance of co-ratings of the one or more items is smaller than a specified threshold.

77. The apparatus as recited in claim 61, wherein identifying the one or more other users includes choosing the one or more other users who have co-rated a subset of the one or more items exactly the same as the particular user.

78. The apparatus as recited in claim 61, wherein identifying the one or more other users includes choosing the one or more other users who have rated co-rated a subset of the one or more items within a specified threshold of the ratings by the particular user.

79. The apparatus as recited in claim 61, wherein identifying the one or more other users includes consideration of one or more of whether the one or more users have a correlation of co-ratings of the one or more items greater than a specified threshold, whether the one or more users have a vector cosine distance of co-ratings of the one or more items smaller than a specified threshold, whether the one or more users have co-rated a subset of the one or more items exactly the same, and whether the one or more users having co-rated a subset of the one or more items within a specified threshold of the ratings by the particular user.

80. The apparatus as recited in claim 61, wherein identifying the one or more other users includes choosing a specified number of other users most similar to the particular user.

81. The apparatus as recited in claim 61, wherein identifying the one or more other users includes choosing other users whose similarity to the particular user is better than a specified threshold.

82. The apparatus as recited in claim 61, wherein identifying the one or more other users includes choosing up to a specified number of other users most similar to the particular user, where all the chosen other users have similarity better than a specified threshold.

83. The apparatus as recited in claim 61, wherein generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users includes determining an average rating of the particular item by the one or more other users.

84. The apparatus as recited in claim 61, wherein generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users includes determining an average rating of the particular item by the one or more other users.

85. The apparatus as recited in claim 61, wherein the memory further comprises one or more additional instructions which, when processed by the one or more processors, cause the one or more processors to perform the additional steps of:

generating, for each of the one or more users, a separate estimation of how the particular user would rate the particular item based upon the rating for the particular item given by each of the one or more other users, and

calculating a confidence metric based upon disparities between the separate estimations.

86. The apparatus as recited in claim 61, wherein the plurality of items is a plurality of movies, the one or more items is one or more movies and the particular item is a particular movie.

87. The apparatus as recited in claim 61, wherein the plurality of items is a plurality of music items, the one or more items is one or more music items and the particular item is a particular music item.

88. An apparatus configured to:

generate an estimation of how a particular user would rate a particular item from a plurality of items by;

identifying, from the plurality of items, one or more items that have ratings similar to the particular item;

identifying one or more other users that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items; and

generating the estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.

89. The apparatus as recited in claim 88, wherein the step of generating the estimation is performed in advance of the need for the estimation and data that indicates the estimation is stored in a memory.

90. The apparatus as recited in claim 89, wherein the apparatus is further configured to write data that indicates the estimation to a non-volatile storage.

91. The apparatus as recited in claim 90, wherein the non-volatile storage is one or more magnetic or solid-state disc systems, all or portions of which are memory-mapped into a random-access memory as needed.

92. The apparatus as recited in claim 90, wherein the apparatus is further configured to retrieve the data that indicates the estimation from the non-volatile storage and write the data into a volatile memory.

93. The apparatus as recited in claim 88, wherein the apparatus is further configured to in response to one or more ratings being updated or added for any of the one or more other items or the particular item:

identifying one or more other items from the plurality of items that have ratings similar to the particular item, wherein the one or more other items may include any of the one or more items;

identifying one or more users that have given ratings to the one or more other items that are substantially similar to ratings given by the particular user to the one or more other items;

generating a revised estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more users; and

storing data that indicates the revised estimation of how the particular user would rate the particular item.

94. The apparatus as recited in claim 88, wherein the apparatus is further configured to in response to one or more ratings being updated or added for any of the one or more other items or the particular item:

identifying one or more estimated ratings that are affected by the one or more ratings being updated or added;

generating one or more updated estimated ratings for the one or more estimated ratings that are affected by the one or more ratings being updated or added; and

storing data that indicates the one or more updated estimated ratings.

95. The apparatus as recited in claim 88, wherein the apparatus is further configured to periodically generate and update similarity data that identifies similarities between items from the plurality of items based upon ratings associated with the plurality of items.

96. The apparatus as recited in claim 88, wherein the apparatus is further configured to select the one or more other users from a plurality of users based upon selection criteria that includes whether a particular user has a specified number of ratings and to generate and store data that indicates the selected one or more other users.

97. The apparatus as recited in claim 88, further comprising one or more prioritized task queues from which tasks are assigned to be processed based upon a priority associated with each task and wherein tasks associated with updating estimated ratings for active users are assigned the highest priority.

98. The apparatus as recited in claim 88, wherein the plurality of items is a plurality of movies, the one or more items is one or more movies and the particular item is a particular movie.

99. The apparatus as recited in claim 88, wherein the plurality of items is a plurality of music items, the one or more items is one or more music items and the particular item is a particular music item.

100. An apparatus comprising:

a plurality of processes assigned to generate a plurality of estimated ratings of how users would rate items from a plurality of items, wherein generating an estimate of how a particular user would rate a particular item from the plurality of items is performed by;

identifying one or more items from the plurality of items that have ratings similar to the particular item;

identifying one or more other users that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items; and

generating the estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.

101. The apparatus as recited in claim 100, further comprising one or more processes that are configured to generate and store ratings data that indicates estimated ratings, wherein each such process is not given access to ratings data generated and stored by other such processes.

102. The apparatus as recited in claim 101, further comprising one or more processes assigned to select reference users and identify similarities between items from the plurality of items based upon ratings associated with the plurality of items and to generate and store similarity data that indicates the identified similarities between items from the plurality of items, wherein the one or more processes have read/write access to the similarity and reference user data and the one or more other processes have read-only access to the similarity data.

103. The apparatus as recited in claim 102, wherein the similarity and reference user data is copied periodically and the one or more recommendation processes have access to the copied similarity and reference user data.

104. The apparatus as recited in claim 100, further comprising an additional process assigned to select the one or more other users from a plurality of users based upon selection criteria that includes whether a user has a specified number of ratings and to generate data that indicates the selected one or more other users, wherein the additional process has read/write access to the data and the plurality of processes have read-only access to the data.

105. The apparatus as recited in claim 100, wherein the plurality of items is a plurality of movies, the one or more items is one or more movies and the particular item is a particular movie.

106. The apparatus as recited in claim 100, wherein the plurality of items is a plurality of music items, the one or more items is one or more music items and the particular item is a particular music item.

107. A machine-implemented method for estimating how a particular user would rate a particular item from a plurality of items, the method comprising the machine-implemented steps of:

identifying one or more items from the plurality of items that are similar to the particular item by associating semantic feature values with the plurality of items and the particular item, where each feature type has a match-score and a mismatch-score, either of which may be positive, zero, or negative, and scoring the similarity of items is performed by

finding the features that have matching values and summing the associated match-score, and

finding the features that have mismatched values and summing the associated mismatch-score,

wherein items with a summed similarity score above a specified threshold are considered similar or the specified number of items with the highest similarity score are considered similar;

identifying one or more other users that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items; and

generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.

108. A machine-readable medium for estimating how a particular user would rate a particular item from a plurality of items, the machine-readable medium carrying instructions which, when processed by one or more processors, cause performance of the steps of:

identifying one or more items from the plurality of items that are similar to the particular item by associating semantic feature values with the plurality of items and the particular item, where each feature type has a match-score and a mismatch-score, either of which may be positive, zero, or negative, and scoring the similarity of items is performed by

finding the features that have matching values and summing the associated match-score, and

finding the features that have mismatched values and summing the associated mismatch-score,

wherein items with a summed similarity score above a specified threshold are considered similar or the specified number of items with the highest similarity score are considered similar;

identifying one or more other users that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items; and

generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.

109. An apparatus for estimating how a particular user would rate a particular item from a plurality of items, the apparatus comprising a memory storing instructions which, when processed by one or more processors, cause performance of the machine-implemented steps of:

identifying one or more items from the plurality of items that are similar to the particular item by associating semantic feature values with the plurality of items and the particular item, where each feature type has a match-score and a mismatch-score, either of which may be positive, zero, or negative, and scoring the similarity of items is performed by

finding the features that have matching values and summing the associated match-score, and

finding the features that have mismatched values and summing the associated mismatch-score,

wherein items with a summed similarity score above a specified threshold are considered similar or the specified number of items with the highest similarity score are considered similar;

identifying one or more other users that have given ratings to the one or more items that are substantially similar to ratings given by the particular user to the one or more items; and

generating an estimation of how the particular user would rate the particular item based upon ratings for the particular item given by the one or more other users.