Person to person similarities based on media experiences

• US 8,301,692 B1
• Filed: 06/16/2009
• Issued: 10/30/2012
• Est. Priority Date: 06/16/2009
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Hamaker, Janna, Kalenkovich, Eugene, Killalea, Tom, Kambarov, Zaur
• Primary Examiner(s): Etienne, Ario
• Assistant Examiner(s): Williams, Clayton R
• Application Number:
  US12/485,700
• Time in Patent Office: 1,232 Days
• Field of Search: None
• US Class Current: 709/203
• CPC Class Codes: G06Q 30/0631 Item recommendations

Abstract

Similarities are discovered among different users with respect to their media experiences and other behaviors, such as taste in media items (e.g., books, music, movies, magazines, art, etc.), browsing behavior, purchase decisions, and online shopping habits, and usage history. The similarities are determined in part by substantially real-time comparison of individual users with a set of predetermined user-based clusters formed from the experiences and behaviors of sample users. Users from a population may then be identified based on similarity metrics. Recommendations for new media items, or other goods/services, may also be made based on choices being made by similar users.

Claims

1. A computer-implemented method, comprising:

forming multiple individual user-based clusters according to media items experienced by individual sample users on whom each of the clusters are based; and

computationally deriving a media experience similarity between a first non-sample user and a second non-sample user with respect to the media items the first and second non-sample users have experienced, the media experience similarity being derived in part based on a comparison of a first indicator and a second indicator, the first indicator indicative of how similar a media experience of the first non-sample user is to a media experience associated with the multiple individual user-based clusters and the second indicator indicative of how similar a media experience of the second non-sample user is to a media experience associated with the multiple individual user-based clusters.

2. The computer-implemented method of claim 1, wherein the media items comprise at least one of text, images, audio, and video.

3. The computer-implemented method of claim 1, wherein the sample users are real humans.

4. The computer-implemented method of claim 1, wherein the sample users are fictitious users computed to provide diversity.

5. The computer-implemented method of claim 1, wherein the media items comprise books, and the media experience similarity comprises a reading similarity with respect to the books that the first and second non-sample users have read.

6. The computer-implemented method of claim 1, wherein the forming of the multiple individual user-based clusters comprises:

selecting a first sample user;

defining a first individual user-based cluster according to a first set of media items experienced by the first sample user;

selecting a second sample user to satisfy a condition that a second set of media items experienced by the second sample user contains the most items not common with the first set of media items experienced by the first sample user; and

defining a second individual user-based cluster according to the second set of media items experienced by the second sample user.

7. The computer-implemented method of claim 6, further comprising expanding the first and second user-based clusters to include other media items that represent variations of a same body of work as the media items experienced by the first and second sample users.

8. The computer-implemented method of claim 6, wherein the selecting of the first sample user comprises choosing a sample user, from among a population of users, who has experienced a least number of media items that still exceeds a threshold number of media items.

9. The computer-implemented method of claim 6, wherein the selecting of the first sample user comprises choosing a sample user, from among a population of users, who has experienced the most media items.

10. The computer-implemented method of claim 6, further comprising repeating the selecting and defining for C sample users, where C is a positive integer greater than zero.

11. The computer-implemented method of claim 1, further comprising pruning certain ones of the multiple user-based clusters.

12. The computer-implemented method of claim 1, wherein

the first and second indicators are determined by mapping the first and second non-sample users into a multi-dimensional space according to how the media items experienced by each of the first and second non-sample users compares with the media items in each of the user-based clusters;

and

the comparison of the first and second indicators comprises at least computing a distance from the first non-sample user to the second non-sample user within the multi-dimensional space.

13. The computer-implemented method of claim 1, further comprising identifying one or more other non-sample users who have one of a highest or lowest media experience similarity as compared to the first non-sample user.

14. The computer-implemented method of claim 13, further comprising presenting, on a user interface, a list of the other non-sample users.

15. The computer-implemented method of claim 13, further comprising recommending media items to the first non-sample user, wherein the media items are selected from a set of the media items consumed by the other non-sample users who have higher media experience similarities.

16. A computer-implemented method, comprising:

forming multiple individual clusters of items experienced by sample users; and

computationally deriving an experience similarity between a first non-sample user and a second non-sample user with respect to a first cluster of items and a second cluster of items experienced by the first and second non-sample users, respectively, the experience similarity being derived in part based on a comparison of a first indicator and a second indicator, the first indicator indicative of how similar the first cluster of items is to the multiple individual clusters of items experienced by sample users and the second indicator indicative of how similar the second cluster of items is to the multiple individual clusters of items experienced by sample users.

17. The computer-implemented method of claim 16, wherein the items comprise media items.

18. The computer-implemented method of claim 16, wherein

the first and second indicators are determined by mapping the first and second non-sample users into a multi-dimensional space according to how the items experienced by each of the first and second non-sample users compares with the items in each of the individual clusters;

and

the comparison of the first and second indicators comprises at least computing a distance from the first non-sample user to the second non-sample user within the multi-dimensional space.

19. A computer-implemented method, comprising:

under control of a computing device configured with executable instructions, performing the following;

selecting, from a population of users, a first sample user who experiences a first set of media items;

defining a first individual user-based cluster according to the first set of media items experienced by the first sample user;

selecting, from the population of users, a second sample user who experiences a second set of media items;

defining a second individual user-based cluster according to the second set of media items experienced by the second sample user; and

repeating the selecting and defining to form C individual user-based clusters, where C is a positive integer greater than two; and

under control of a computing device configured with executable instructions, performing the following in substantially real time;

receiving a query from a non-sample user; and

determining a media experience similarity between the non-sample user and one or more other non-sample users with respect to the media items the non-sample users have experienced, the media experience similarity between the non-sample user and each of the other non-sample users being derived in part based on a comparison of a first indicator and a second indicator, the first indicator indicative of how similar a media experience of the non-sample user is to a media experience associated with each of the individual user-based clusters and the second indicator indicative of how similar a media experience of the other non-sample user is to a media experience associated with each of individual user-based cluster.

20. The computer-implemented method of claim 19, wherein the selecting a first sample user comprises selecting a first sample user who has experienced a least number of media items that still exceeds a threshold number of media items.

21. The computer-implemented method of claim 20, wherein the selecting a second sample user comprises selecting a second sample user who has experienced a next least number of media items that still exceeds the threshold number and additionally meets a criteria that the second set of media items contains most items not common with the first set the media items.

22. The computer-implemented method of claim 21, further comprising excluding media items in the first set of media items from a universe of media items to form a smaller universe of media items before selecting the second sample user, wherein the selecting of the second sample user comprises selecting a second sample user who has experienced a least number of media items in the smaller universe that still exceeds a threshold number of media items.

23. The computer-implemented method of claim 19, further comprising suggesting non-sample users of interest based on the media experience similarity.

24. The computer-implemented method of claim 19, further comprising recommending media items that are experienced by other non-sample users who are determined to have similar media experiences.

25. A non-transitory computer-readable media storing computer-readable instructions that, when executed, instruct a processor to perform acts comprising:

forming multiple individual user-based clusters according to behaviors exhibited by individual sample users on whom each of the clusters are based;

mapping the user-based clusters as orthogonal vectors in a multi\-dimensional space;

placing a first non-sample user into the multi-dimensional space according to how the media items experienced by the first non-sample user compares with the media items in each of the individual user-based clusters;

placing a second non-sample user into the multi-dimensional space according to how the media items experienced by the second non-sample user compares with the media items in each of the individual user-based clusters; and

computing a similarity metric between the first and second non-sample users as a function of a distance from the first non-sample user to the second non-sample user within the multi-dimensional space.

26. The non-transitory computer-readable media of claim 25, wherein the behaviors comprise at least one of purchase decisions, browsing behavior, media items experienced, and usage history.

27. The non-transitory computer-readable media of claim 25, wherein the forming multiple user-based clusters comprises:

selecting a first sample user who exhibits a first set of behaviors;

defining a first individual user-based cluster according to the first set of behaviors;

selecting a second sample user who exhibits a second set of behaviors, the second sample user being selected to satisfy a condition that the second set of behaviors intersect least with the first set of behaviors; and

defining a second individual user-based cluster according to the second set of behaviors.

28. The non-transitory computer-readable media of claim 27, wherein the forming multiple individual user-based clusters further comprises repeating the selecting and defining for C sample users, where C is a positive integer greater than two.

29. The non-transitory computer-readable media of claim 25, further comprising pruning certain ones of the multiple individual user-based clusters.

30. The non-transitory computer-readable media of claim 25, further comprising ranking non-sample users relative to the first non-sample user according to similarity metrics.

31. The non-transitory computer-readable media of claim 25, further comprising presenting, via a user interface, a list of non-sample users relative to the first non-sample user and ordering the list according to similarity metrics.

32. The non-transitory computer-readable media of claim 25, further comprising presenting, via a user interface, a list of non-sample users relative to the first non-sample user and enabling the first non-sample user to elect to follow one or more of the non-sample users with respect to the behaviors exhibited.

33. The non-transitory computer-readable media of claim 25, further comprising making recommendations to the first non-sample user based on behaviors observed of one or more other non-sample users with relatively high similarity metrics.