Discovery of behavior-based item relationships based on browsing session records

• US 8,407,105 B2
• Filed: 06/06/2011
• Issued: 03/26/2013
• Est. Priority Date: 09/18/1998
• Status: Expired due to Fees

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon.com, Inc.
• Inventors: Linden, Gregory D., Smith, Brent R., Zada, Nida K.
• Primary Examiner(s): ZARE, SCOTT A
• Application Number:
  US13/154,334
• Publication Number:
  US 20110238525A1
• Time in Patent Office: 659 Days
• Field of Search: 705/7.29, 705/14.53, 705/14.73, 705/26.7, 705/26.8, 705/27.1
• US Class Current: 705/26.7
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0253 During e-commerce, i.e. online transactions
• CPC Class Codes: G06Q 30/0254 based on statistics
• CPC Class Codes: G06Q 30/0255 based on user history
• CPC Class Codes: G06Q 30/0256 User search
• CPC Class Codes: G06Q 30/0271 Personalized advertisement
• CPC Class Codes: G06Q 30/0277 Online advertisement
• CPC Class Codes: G06Q 30/0282 Rating or review of business operators or products
• CPC Class Codes: G06Q 30/0601 Electronic shopping [e-shopping]
• CPC Class Codes: G06Q 30/0603 Catalogue ordering
• CPC Class Codes: G06Q 30/0631 Item recommendations
• CPC Class Codes: G06Q 30/0633 Lists, e.g. purchase orders, compilation or processing
• CPC Class Codes: G06Q 30/0641 Shopping interfaces
• CPC Class Codes: H04L 51/02 using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface

Abstract

Various processes are disclosed for discovering item relationships between particular items, such as products represented in an electronic catalog, based on monitored user behaviors (e.g., item viewing activities, item purchases, shopping cart activities, etc.). The discovered item relationships may, for example, be used to generate personalized item recommendations for users, and/or to supplement item detail pages of an electronic catalog with lists of related items. Also disclosed are processes for generating personalized item recommendations based on users'"'"' search activities and browse node visits.

Claims

1. A computer-implemented method of discovering behavior-based relationships between items represented in an electronic catalog, the method comprising:

recording accesses by users to item detail pages of the electronic catalog, each item detail page corresponding to a respective catalog item, said accesses recorded without requiring the users to rate or purchase the items and without requiring the users to create lists of the items;

storing a plurality of session records in computer storage, each session record being specific to a respective browsing session of a user and identifying a plurality of items for which the user accessed a corresponding item detail page during the browsing session, at least some of said session records corresponding to different users than other session records;

generating data values representing strengths of relationships between particular ones of said items, wherein generating the data values comprises, for a pair of items consisting of a first item and a second item, determining, based on the plurality of session records, at least (1) the number of browsing sessions in which an item detail page for the first item was accessed, (2) the number of browsing sessions in which an item detail page for the second item was accessed, and (3) the number of browsing sessions in which item detail pages for both the first item and the second item were accessed; and

generating a mapping of items to related items based at least partly on said data values;

said method performed by a computing system that comprises one or more computers, wherein the method enables machine-based discovery of behavior-based item relationships without requiring the users to rate or purchase the items and without requiring the users to create lists of the items.

2. The method of claim 1, wherein generating the data values comprises generating, for said pair of items, a data value based on the equation N_common/sqrt(N_A×

N_B), where N_common is the number of browsing sessions in which item detail pages for both the first item and the second item were accessed, sqrt is a square-root operation, N_A is the number of browsing sessions in which an item detail page for the first item was accessed, and N_B is the number of browsing sessions in which an item detail page for the second item was accessed.

3. The method of claim 1, further comprising using said mapping to supplement item detail pages of the electronic catalog with lists of related items.

4. The method of claim 1, further comprising causing an item detail page for a first item to be supplemented with a list of additional items and with an indication that the additional items are viewed by users who view the first item, said list based on said mapping.

5. The method of claim 1, further comprising using the mapping, in combination with a record of a plurality of items selected by a user during a current browsing session, to select items to recommend to the user during the current browsing session.

6. The method of claim 1, wherein generating the data values comprises using item purchase histories of users as an additional measure of item relatedness.

7. The method of claim 1, further comprising generating said session records based on monitored browsing activity of users such that each session record corresponds to a respective period of substantially continuous browsing activity.

8. The method of claim 1, further comprising generating said session records based on monitored browsing activity of users, wherein generating said session records comprises determining whether a particular browsing session has ended based at least partly on whether a checkout transaction has occurred.

9. The method of claim 1, wherein at least some of the session records are stored as anonymous session records that are not linked to corresponding users.

10. A computer-implemented method of discovering behavior-based relationships between items represented in an electronic catalog, the method comprising:

storing a plurality of session records in computer storage, each session record being specific to a respective browsing session of a user and identifying the items selected by the user for viewing in the electronic catalog during the browsing session, at least some of said session records corresponding to different users than other session records;

generating data values representing strengths of relationships between particular ones of said items, wherein generating the data values comprises, for a pair of items consisting of a first item and a second item, determining, based on the plurality of session records, the number of browsing sessions in which both items of the pair were selected for viewing and the number of browsing sessions in which the first item was selected for viewing; and

generating a mapping of items to related items based at least partly on said data values;

said method performed by a computing system that comprises one or more computers.

11. A system for discovering behavior-based relationships between items represented in an electronic catalog, the system comprising:

a computer data repository that stores a plurality of session records, each session record being specific to a respective browsing session of a user and identifying items selected by the user for viewing in an electronic catalog during the browsing session, at least some of said session records corresponding to different users than other session records, said session records generated without requiring the users to rate or purchase the items and without requiring the users to create lists of the items; and

a computing system programmed to at least;

generate data values representing strengths of relationships between particular ones of said items using a method that comprises, for an item pair consisting of a first item and a second item, determining, based on the plurality of session records, at least (1) the number of sessions in which the first item was selected for viewing, (2) the number of sessions in which the second item was selected for viewing, and (3) the number of sessions in which both items of the pair were selected for viewing; and

generate a mapping of items to related items based at least partly on said data values;

whereby the system enables behavior-based relationships between items to be discovered without requiring the users to rate or purchase the items and without requiring the users to create lists of the items.

12. The system of claim 11, wherein the system, in generating the data values, is configured to treat each user access to an item detail page of the electronic catalog as a selection of a corresponding item for viewing.

13. The system of claim 12, further comprising a component that is configured to use said mapping to supplement item detail pages of the electronic catalog with corresponding lists of related items.

14. The system of claim 11, further comprising a component that is configured to supplement an item detail page for the first item with a list of additional items, and with an indication that the additional items are viewed by users who view the first item, said list based on the mapping.

15. The system of claim 11, further comprising a recommendation system configured to use the mapping, in combination with a record of a plurality of items selected by a user for viewing during a current browsing session, to select items to recommend to the user during the current browsing session.

16. The system of claim 11, further comprising a component that is configured to generate said session records based on monitored browsing activity of users such that each session record corresponds to a respective period of substantially continuous browsing activity.

17. The system of claim 11, further comprising a component that is configured to generate the session records based on monitored browsing activity of users, said component configured to determine whether a particular browsing session has ended based at least partly on whether a checkout transaction has occurred.

18. The system of claim 11, wherein at least some of the session records are stored in said data repository as anonymous session records.

19. A system for discovering behavior-based relationships between items represented in an electronic catalog, the system comprising:

a computer data repository that stores a plurality of session records, each session record being specific to a respective browsing session of a user and identifying the items selected by the user for viewing in the electronic catalog during the browsing session, at least some of said session records corresponding to different users than other session records; and

a computing system programmed to at least;

generate data values representing strengths of relationships between particular ones of said items using a method that comprises determining, for a pair of items consisting of a first item and a second item, based on the plurality of session records, the number of sessions in which both items of the pair were selected for viewing and the number of sessions in which the first item of the pair was selected for viewing; and

generate a mapping of items to related items based at least partly on said data values.

20. The system of claim 19, wherein each session record identifies a plurality of items selected by the respective user for viewing in the electronic catalog during the browsing session.

21. A system for discovering and exposing behavior-based relationships between items represented in an electronic catalog, the system comprising:

a monitoring system that is configured to generate records of item viewing activities of a plurality of users, said records of item viewing activities identifying particular items selected by users for viewing in an electronic catalog of items, said monitoring system configured to generate said records without requiring the users to rate or purchase the items and without requiring the users to create lists of the items;

a first component that is configured to generate a mapping of items to related items based on the records of item viewing activities, said mapping reflecting item-viewing-based relationships between items, said first component comprising computer hardware, said first component configured to measure relationships between items by a process that comprises, for an item pair consisting of a first item and a second item, determining at least (1) the number of sessions in which the first item was selected for viewing, (2) the number of sessions in which the second item was selected for viewing, and (3) the number of sessions in which both the first item and the second item were selected for viewing; and

a second component that is configured to use the mapping to incorporate lists of related items into item detail pages of the electronic catalog;

said system enabling behavior-based item relationships to be discovered without requiring the users to rate or purchase the items and without requiring the users to create lists of the items.

22. The system of claim 21, wherein the second component is configured to supplement an item detail page for the first item with a list of additional items, and with an indication that the additional items are viewed by users who view the first item, said list based on said mapping.

23. The system of claim 21, wherein the first component is configured to generate, for the item pair, a data value based on the equation N_common/sqrt(N_A×

N_B), where N_common is the number of sessions in which both the first item and the second item were selected for viewing, sqrt is a square-root operation, N_A is the number of sessions in which the first item was selected for viewing, and N_B is the number of sessions in which the second item was selected for viewing.

24. The system of claim 21, wherein the items are products capable of being purchased via the electronic catalog.

25. The system of claim 21, wherein the records of item viewing activities are based on visits to item detail pages of the electronic catalog.