Systems and methods for statistically selecting content items to be used in a dynamically-generated display

• US 7,594,189 B1
• Filed: 03/29/2006
• Issued: 09/22/2009
• Est. Priority Date: 04/21/2005
• Status: Expired due to Fees

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Walker, Toby H., Sundberg, Patrik P., Rauser, John M., Selinger, David L.
• Primary Examiner(s): Zhou; Ting
• Assistant Examiner(s): ZAHR, ASHRAF A
• Application Number:
  US11/392,158
• Time in Patent Office: 1,273 Days
• Field of Search: 705 26- 27, 715/811
• US Class Current: 715/811
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

An apparatus and methods advantageously select content items for dynamically-generated web pages in an intelligent and virtually autonomous manner. This permits the operator of the web site to rapidly identify and respond to trends, thereby advantageously updating the web site relatively quickly and efficiently without or with less time consuming and expensive manual labor. User interaction for a plurality of users with the web site is collected in a database. For various content items, the database is mined to extract relationships between probability and references of select attributes in probability models. When a new web page is requested, attributes, which can include attributes associated with a user, are used as references to the applicable probability models of selected content items, combined with value weighting to generate expected values, and selected for use in the web page at least partially based on the expected values.

Claims

1. A system for selecting one or more content items in a dynamically-generated display, the system comprising:

a web server computer system comprising a memory operatively coupled to a processor, the web server computer system configured to receive a request for a display from a client computer, and to relate the client computer to a particular user; and

a content selection computer system in communication with the web server computer system, wherein the content selection computer system is configured to retrieve values for a plurality of attributes of the particular user, and, for each content item of a plurality of content items, to;

retrieve a probability model associated with the content item;

compute probabilities with the retrieved attribute values for the particular user as inputs for the probability model;

retrieve payoff values associated with the one or more user responses to the content item; and

retrieve confidence intervals for the payoff values for each content item, wherein a confidence interval is a range of the payoff values bounded by a confidence factor;

wherein the content selection computer system is configured to select a content item from the plurality of content items for the dynamically-generated display based at least on the probabilities, payoff values, and confidence intervals.

2. The system as defined in claim 1, wherein the content selection computer system is configured to select the content item based on the probabilities, payoff values, and confidence intervals for the content items.

3. The system as defined in claim 1, wherein the display comprises a web page.

4. The system as defined in claim 1, wherein the probability models have attribute ranges for a plurality of attributes that are segmented into a plurality of segments, wherein the particular user'"'"'s attribute values maps to one segment of the plurality.

5. The system as defined in claim 1, wherein the content selection computer system is configured to select a first content item with equal probability as a second content item when their confidence intervals overlap.

6. A method of statistically selecting one or more content items for a dynamically-generated display, the method comprising:

retrieving a plurality of probability models associated with candidate content items;

using the probability models to compute probabilities of an occurrence of one or more interactive user responses to exposure to the candidate content items;

retrieving one or more payoffs associated with the one or more interactive user responses to the candidate content items and one or more confidence intervals associated with the one or more payoffs, wherein a confidence interval is a range associated with the payoffs bounded by a confidence factor; and

selecting one or more content items from the candidate content items for the dynamically-generated display, wherein the selection is made automatically by a computer system and is based at least on the computed probabilities, the retrieved payoffs, and the confidence intervals.

7. The method as defined in claim 6, further comprising using the probability models to dynamically generate confidence intervals of expected values for content items of a first list, and selecting a second list of a plurality of content items for display from the first list based on the confidence intervals.

8. The method as defined in claim 6, further comprising selecting based on the computed probabilities, the retrieved payoffs, and the confidence.

9. The method as defined in claim 6, wherein selecting further comprises:

computing expected values for the one or more user responses for the content items using the probabilities and the payoffs associated with the one or more user responses; and

selecting the one or more content items based on the computed expected values.

10. The method as defined in claim 9, further comprising computing the expected values with a randomization component such that the content items with an otherwise higher expected value are merely more likely to be selected than content items with an otherwise lower expected value.

11. The method as defined in claim 9, further comprising:

computing confidence intervals when applicable or retrieving default confidence intervals for the content items; and

selecting the one or more content items at least partially based on the computed expected values, the computed confidence intervals, the retrieved default expected values, and the retrieved default confidence intervals.

12. The method as defined in claim 9, wherein selecting further comprises selecting based on the highest expected value.

13. The method as defined in claim 6, further comprising:

computing expected values for the one or more user responses to the content items using the probabilities and the values associated with the one or more user responses;

retrieving default expected values for content items for which a usable probability model is not available; and

selecting the one or more content items at least partially based on the computed expected values and the retrieved default expected values.

14. The method as defined in claim 13, further comprising:

computing confidence intervals when applicable or retrieving default confidence intervals for the content items; and

selecting the one or more content items at least partially based on the computed expected values, the computed confidence intervals, the retrieved default expected values, and the retrieved default values of confidence intervals.

15. The method as defined in claim 6, further comprising:

retrieving predetermined values for at least a portion of the content items for at least one of the probability of a user response or a computation based on the probability; and

selecting the one or more content items using calculated probabilities and predetermined probabilities.

16. The method as defined in claim 6, when a probability model is not available for a content item, using one of manual entry, copying from default data, or copying from data from a similar content item to substitute for the probability model, and selecting the one or more content items using probabilities calculated from probability models and substitutes.

17. The method as defined in claim 16, further comprising:

computing confidence intervals when applicable or retrieving default confidence intervals for the content items; and

selecting the one or more content items at least partially based on the computed expected values, the computed confidence intervals, the retrieved default expected values, and the retrieved default confidence intervals.

18. The method as defined in claim 6, further comprising retrieving attributes associated with a user of the display.

19. A method for generating a plurality of probability models for a content item that can be displayed in an interactive display, the method comprising:

(a) selecting the content item for which the plurality of probability models are to be generated;

(b) collecting records of impressions relevant to the content item;

(c) for each relevant impression, performing (d), (e), and (f) to collect information;

(d) determining zero or more actual user responses for the impression;

(e) retrieving attribute values corresponding to the impression;

(f) associating the retrieved attribute values with the zero or more user responses to the impression;

(g) retrieving collections of ranges for a plurality of attributes;

(h) determining individual segmented probability sub-models for each type of applicable user response and applicable collection of ranges for the plurality of attributes based on the associated retrieved attribute values and user responses; and

(i) computing a confidence interval of an expected value for each probability sub-model, wherein a confidence interval is a range of expected value bounded by a confidence factor;

wherein the method is performed automatically by a computer-implemented system.

20. The method as defined in claim 19, wherein the content item corresponds to a good offered for sale through a web site for the web page.

21. The method as defined in claim 19, wherein the impression includes displaying and at least one of no interaction, interaction by clicking, interaction with a web page by selecting, interaction with a game, or interaction associated with viewing a video clip.

22. The method as defined in claim 19, where the actual user response corresponds to at least one of no response, purchasing a related item, adding to an electronic shopping cart, adding to a gift basket, or adding to a wish list.

23. The method as defined in claim 19, further comprising identifying those attributes which are correlated with user actions, collecting attribute values for user responses and determining individual probability sub-models for only those attributes.

24. The method as defined in claim 19, wherein the confidence interval is computed based on an estimated sample variance and on a selected confidence factor, wherein the confidence factor corresponds to a degree to which an actual value is expected to fall within the confidence interval.