Detection of behavior-based associations between search strings and items

• US 7,953,740 B1
• Filed: 02/13/2006
• Issued: 05/31/2011
• Est. Priority Date: 02/13/2006
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Redman, Benjamin W. S., Kohavi, Ron, Jayaraman, Gautam K., Whitman, Ronald M., Vadon, Eric R.
• Primary Examiner(s): Vu; Bai D
• Application Number:
  US11/276,079
• Time in Patent Office: 1,933 Days
• Field of Search: None
• US Class Current: 707/751
• CPC Class Codes: G06F 16/337 Profile generation, learning or modification
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

A system and method are disclosed for automatically detecting associations between particular sets of search criteria, such as particular search strings, and particular items. Actions of users of an interactive system, such as a web site, are monitored over time to generate event histories reflective of searches, item selection actions, and possibly other types of user actions. An analysis component collectively analyzes the event histories to automatically identify and quantify associations between specific search strings (or other types of search criteria) and specific items. As part of this process, a decay function reduces the weight given to a post-search item selection event based on intervening events that occur between the search event and the item selection event.

Claims

1. A computer-implemented method of measuring a degree of association between a search string and a selectable item, the method comprising:

identifying a search string submission event in which a user submitted a search string to search an electronic repository;

identifying an item selection event in which the user selected an item represented in said electronic repository, said item selection event occurring after the search string submission event;

identifying one or more intervening events initiated by the user subsequent to the search string submission event but prior to the item selection event, wherein the item selection event, the one or more intervening events and the search string submission event comprise a user event history sequence;

calculating a weight value that represents a degree of association between the search string submission event and the item selection event, wherein calculating the weight value comprises assigning different decay values to different ones of said one or more intervening events from the user event history sequence, wherein the decay values are dependent upon a type of said intervening event, and using the decay values in combination to calculate said weight value; and

using said weight value, in combination with weight values corresponding to other event sequences, to calculate a relationship value representing a degree to which the search string and the item are associated.

2. The method of claim 1, wherein the weight value specifies an amount of weight given to the item selection event in calculating said relationship value.

3. The method of claim 1, wherein calculating the weight value comprises taking into consideration the number of said one or more intervening events that occurred between the search string submission event and the item selection event.

4. The method of claim 1, wherein calculating the weight value comprises taking into consideration a type of an intervening event that occurred between the search string submission event and the item selection event.

5. The method of claim 1, wherein calculating the weight value comprises decreasing an initial weight value for each of the one or more intervening events, such that each intervening event serves to reduce an amount of weight given to the item selection event for purposes of calculating the relationship value.

6. The method of claim 1, wherein the one or more intervening events includes a second search string submission event, and wherein calculating said weight value comprises taking into consideration a degree to which respective search strings of the first and second search string submission events are related.

7. The method of claim 1, wherein calculating the weight value comprises taking into consideration the amount of time that passed between the search string submission event and the item selection event.

8. The method of claim 1, wherein the item is a product represented in an electronic catalog.

9. The method of claim 1, wherein the item is a category of products represented in an electronic catalog.

10. The method of claim 1, wherein the item is a web site, and the item selection event is an event in which the user selects the web site from a search results page.

11. The method of claim 1, wherein the other event sequences include at least one event sequence corresponding to another user.

12. A computer system, comprising:

a data repository that stores event histories of a plurality of users of a system that provides interactive functionality for users to search for and select items, said event histories including at least search events and item selection events; and

a computing device in communication with the data repository that executes an analysis component, the analysis component quantifying associations between particular search strings and particular items, at least in part, by analyzing the event histories of the plurality of users;

wherein the analysis component is configured to calculate, for a user event history sequence that comprises a search event followed by one or more intervening events followed by an item selection event, an event weight that specifies an amount of weight to give to the item selection event for purposes of quantifying an association between a search string submitted in said search event and an item selected in the item selection event, wherein the analysis component calculates the event weight as a function of a decay value assigned to each of the one or more intervening events in the user event history sequence, and as a function of at least one of the following in the user event history sequence;

(1) a number of said one or more intervening events between the search event and the item selection event, (2) a type of at least one of the one or more intervening events, wherein the analysis component assigns different non-zero decay values to different types of intervening events.

13. The system of claim 12, wherein the system is configured to use said weight, in combination with event weights calculated for other item selection events associated with said item and said search string, to generate a relationship value representing a degree of association between the search string and the item.

14. The system of claim 12, wherein the decay value assigned to each of the one or more intervening events serves to reduce said amount of weight given to the item selection event.

15. The system of claim 12, wherein the analysis component is configured to assign the decay value to each of the one or more intervening events.

16. The system of claim 15, wherein, when the one of more intervening events comprises an intervening search event, the analysis component assigns to the intervening search event a decay value that is dependent upon a degree of relationship between the said search string and a search string submitted in the intervening search event.

17. The system of claim 12, wherein the analysis component additionally calculates the event weight as a function of the amount of time between the search event and the item selection event.

18. The system of claim 12, wherein the analysis component calculates the event weight as function of both (1) the number of said one or more intervening events between the search event and the item selection event, (2) the type of at least of the one or more intervening events.

19. The system of claim 12, wherein the items are products represented in a browsable electronic catalog.

20. The system of claim 12, wherein the items are web sites accessed via an Internet search engine.

21. A computer-readable storage medium which stores a computer program comprising:

an identification component that identifies a user event history sequence, wherein the user event history sequence comprises;

a search string submission event in which a user submitted a search string to search an electronic repository;

an item selection event in which the user selected an item represented in said electronic repository, said item selection event occurring after the search string submission event;

one or more intervening events initiated by the user subsequent to the search string submission event but prior to the item selection event;

a calculation component that calculates a value that represents a degree of association between the search string submission event and the item selection event based at least in part on a decay function related to each of the one or more intervening events in the user event history sequence, wherein said value is dependent upon a type of the intervening event; and

an association quantification component that takes said value into consideration in quantifying a degree to which the search string and the item are associated, wherein the decay function related to each of the one or more intervening events serves to reduce an amount of weight given to the item selection event for purposes of quantifying said association.

22. The computer-readable medium of claim 21, wherein the association quantification component uses said value, in combination with values corresponding to other item selection events involving the item and the search string, to quantify said degree of association.

23. The computer-readable medium of claim 21, wherein the computer program further comprises a search engine component that uses the quantified degree of association between the search string and the item to determine whether to augment a query result set with said item.

24. The computer-readable medium of claim 23, wherein the search engine component inhibits the item from being added to the query result set if a popularity level of the item exceeds a selected threshold.