System and method for adaptively selecting and delivering recommendations to a requester

• US 7,860,925 B1
• Filed: 05/03/2007
• Issued: 12/28/2010
• Est. Priority Date: 10/19/2001
• Status: Active Grant

• Current Assignee: Outlooksoft Corporation (SAP SE)
• Original Assignee: Outlooksoft Corporation (SAP SE)
• Inventors: Mayer, James, Lou, Qingwei, Gheorghe, Christian
• Primary Examiner(s): Hamza; Faruk
• Application Number:
  US11/800,049
• Time in Patent Office: 1,335 Days
• Field of Search: 709/200, 709/203, 709223-224, 709/201, 707 3- 6, 707/10
• US Class Current: 709/203
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99936 Pattern matching access

Abstract

A system and method for employing a number of producer modules to produce and deliver recommendations to a requester. Each of the recommendations having associated therewith a confidence level. Each of the producer modules having associated therewith a weighting value. The confidence levels in each of the recommendations produced by the producers being modified based on weighting value associated with the producer that produced the produced recommendation. The weighing values associated with each of the producer modules being modified based on information from the requester related to the recommendations delivered to the requester.

Claims

1. A recommendation system for delivering a list of recommendations to a requester system over a computer network, the recommendation system comprising:

one or more processing units executing instructions implementing;

a plurality of producer modules in communication with the requester system, each of the plurality of producer modules receiving a request for recommendations from the requester system and producing a list of initial recommendations in response thereto, each of the recommendations in a list of initial recommendations including a confidence level and a producer identifier; and

a recommendation engine configured to;

modify each of the confidence levels in a given list of initial recommendations based on a weighting value associated with the producer module that produced the given list of initial recommendations; and

select one or more of the recommendations from the lists of initial recommendations based on the confidence levels of the recommendations to produce a list of survived recommendations that are transmitted to the requester system;

wherein the recommendation engine is further configured to;

combine the lists of initial recommendations into a combined list of recommendations;

sort the recommendations in the combined list of recommendations according to the confidence levels of each of the recommendations in the combined list of recommendations to produce a sorted list of recommendations; and

select one or more survived recommendations from the sorted list of recommendations to produce the list of survived recommendations; and

wherein the recommendation engine is further configured to;

select a first predetermined number N of recommendations from the sorted list of recommendations;

select a second predetermined number M of recommendations from the first predetermined number N of recommendations, wherein a likelihood of a given one of the M recommendations being selected from the first predetermined number N of recommendations is related to the confidence level of the given one of the M recommendation; and

select the one or more survived recommendations from the second predetermined number M of initial recommendations.

2. The recommendation system of claim 1, wherein the recommendation engine is further configured to adjust the weighting values associated with each of the producer modules based on information from the requester system related to the list of survived recommendations transmitted to the requester system.

3. The recommendation system of claim 1, wherein the first predetermined number N is characterized by the equation N=2*i+1, wherein i is a predetermined number.

4. The recommendation system of claim 3, wherein i is a number of recommendations to be transmitted in the list of survived recommendations, wherein i is transmitted to the recommendation system from the requester system, and wherein M equals i.

5. The recommendation system of claim 1, wherein the recommendation engine is further configured to modify each of the confidence levels in a given list of initial recommendations by multiplying each of the confidence levels in a given list of initial recommendations by the weighting value associated with the producer module that produced the given list of initial recommendations.

6. The recommendation system of claim 1, wherein at least one of the recommendations in one of the lists of initial recommendations also includes a uniform resource locator (URL).

7. The recommendation system of claim 1, wherein the first predetermined number N of recommendations from the sorted list of recommendations have the first predetermined number N highest confidence levels in the sorted list of recommendations.

8. The recommendation system of claim 7, wherein the recommendation engine is further configured to select the second predetermined number M of recommendations with a degree of randomness, wherein a likelihood of a given one of the M recommendations being selected from the first predetermined number N of recommendations is related to the confidence level of the given one of the M recommendation.

9. The recommendation system of claim 8, wherein the list of initial recommendations produced by one of the plurality of producer modules in response to the request includes a plurality of recommendations;

wherein the confidence level of each of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request is different than the confidence level of each of the others of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request; and

wherein the recommendation engine is configured to modify the different confidence level of each of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request based on the weighting value associated with the one of the plurality of producer modules.

10. The recommendation system of claim 8, wherein for each one of the plurality of producer modules, the list of initial recommendations produced by the one of the plurality of producer modules in response to the request includes a plurality of recommendations;

wherein the confidence level of each of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request is different than the confidence level of each of the others of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request; and

wherein the recommendation engine is configured to modify the different confidence level of each of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request based on the weighting value associated with the one of the plurality of producer modules.

11. The recommendation system of claim 1, wherein the recommendation engine is further configured to select the second predetermined number M of recommendations with a degree of randomness, wherein a likelihood of a given one of the M recommendations being selected from the first predetermined number N of recommendations is related to the confidence level of the given one of the M recommendation.

12. The recommendation system of claim 1, wherein the list of initial recommendations produced by one of the plurality of producer modules in response to the request includes a plurality of recommendations;

wherein the confidence level of each of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request is different than the confidence level of each of the others of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request; and

wherein the recommendation engine is configured to modify the different confidence level of each of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request based on the weighting value associated with the one of the plurality of producer modules.

13. The recommendation system of claim 1, wherein for each one of the plurality of producer modules, the list of initial recommendations produced by the one of the plurality of producer modules in response to the request includes a plurality of recommendations;

wherein the confidence level of each of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request is different than the confidence level of each of the others of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request; and

wherein the recommendation engine is configured to modify the different confidence level of each of the plurality of recommendations in the list of initial recommendations produced by the one of the plurality of producer modules in response to the request based on the weighting value associated with the one of the plurality of producer modules.

14. A relevant object determination system for delivering relevant objects over a computer network to a requester system having one or more users in contact therewith, comprising:

one or more processing units executing instructions implementing;

a first producer module and a second producer module, each of the first and second producer modules receiving a request for relevant recommendations from the requester system, the first producer module producing a first list of recommendations in response to the request for relevant recommendations and the second producer module producing a second list of recommendations in response to the request for relevant recommendations, each recommendation including an object identifier, a confidence level, and a producer identifier; and

a recommendation engine configured to;

modify the confidence level in each recommendation in the first list of recommendations based on a first weighting value and modify the confidence level in each recommendation in the second list of recommendations based on a second weighting value; and

select a predetermined number of the recommendations and deliver to the requester system objects identified by the object identifiers in the predetermined number of recommendations and the producer identifiers in the predetermined number of recommendation;

wherein the recommendation engine is further configured to;

combine the first list of recommendations and the second list of recommendations into a combined list of recommendations;

sort the combined list of recommendations to produce a sorted list of recommendations;

and

select from the sorted list of recommendations the predetermined number of recommendations;

wherein the recommendation engine is further configured to sort the recommendations in the combined list of recommendations according to confidence levels; and

wherein the recommendation engine is further configured to first select N recommendations from the sorted list of recommendations and then select the predetermined number of recommendations from the N selected recommendations, wherein the likelihood of a given one of the recommendations of the predetermined number of recommendations being selected from the N recommendations is related to the confidence level of the given one of the recommendations.

15. The relevant object determination system of claim 14, wherein the recommendation engine is further configured to adjust the first weighting value and the second weighting value based on inputs from the requester system indicative of reactions of the one or more users to objects delivered by the recommendation engine to the requester system.

16. The relevant object determination system of claim 14, wherein the recommendation engine is further configured to:

modify the confidence level in each recommendation in the first list of recommendations by multiplying the confidence level by the first weighting value; and

modify the confidence level in each recommendation in the second list of recommendations by multiplying the confidence level by the second weighting value.

17. The relevant object determination system of claim 14, wherein N is characterized by the equation N=2*i+1, and wherein i is a predetermined number.

18. A method of adaptively weighing producer modules in a recommendation system employing a plurality of producer modules, each producer module having a weighting value associated therewith, the method comprising:

receiving, by one or more processing units, one or more recommendations from each of the plurality of producer modules executing on one or more processing units, each of the received recommendations including a producer identifier indicating which producer module produced the received recommendation and a confidence level;

modifying, by the one or more processing units, the confidence level for each of the received recommendations based on the weighting value associated with the producer module that produced the recommendation;

combining the one or more recommendations from each of the plurality of producer modules into a combined list of recommendations;

sorting the recommendations in the combined list of recommendations according to the confidence levels for each of the received recommendations;

selecting a first predetermined number N of recommendations from the sorted list of recommendations;

selecting a second predetermined number M of recommendations from the first predetermined number N of recommendations, wherein a likelihood of a given one of the M recommendations being selected from the first predetermined number N of recommendations is related to the confidence level of the given one of the M recommendation;

selecting a plurality of survived recommendations from the second predetermined number M of recommendations; and

transmitting the plurality of survived recommendations to a requester system executing on one or more processing units.

19. The method of claim 18, further comprising modifying the confidence level for each of the received recommendations by multiplying each confidence level by the weighting value associated with the producer module that produced the recommendation.

20. The method of claim 18, further comprising:

receiving information from the requester system related to the plurality of survived recommendations transmitted to the requester system; and

modifying each of the weighting values based on the information received from the requester system.

21. The method of claim 20, wherein the information received from the requester system includes a plurality of user reaction values, wherein each of the plurality of user reaction values is associated with a different one of the plurality of survived recommendations transmitted to the requester system.

22. The method of claim 21, wherein each user reaction value is indicative of a positive user reaction to the recommendation to which the user reaction value is associated.

23. A computer readable medium including a storage medium storing computer executable instructions for adaptively weighing producer modules in a recommendation system employing a plurality of producer modules, each producer module having a weighting value associated therewith, the executable instructions comprising instructions executable to result in:

in response to receiving one or more recommendations from each of the plurality of producer modules executing on one or more processing units, each of the received recommendations including a producer identifier indicating which producer module produced the received recommendation and a confidence level, modifying the confidence level for each of the received recommendations based on the weighting value associated with the producer module that produced the recommendation;

combining the one or more recommendations from each of the plurality of producer modules into a combined list of recommendations;

sorting the recommendations in the combined list of recommendations according to the confidence levels for each of the received recommendations;

selecting a first predetermined number N of recommendations from the sorted list of recommendations;

selecting a second predetermined number M of recommendations from the first predetermined number N of recommendations, wherein a likelihood of a given one of the M recommendations being selected from the first predetermined number N of recommendations is related to the confidence level of the given one of the M recommendation;

selecting a plurality of survived recommendations from the second predetermined number M of recommendations; and

transmitting the plurality of survived recommendations to a requester system executing on one or more processing units.

24. The computer readable medium of claim 23, the executable instructions further comprising modifying the confidence level for each of the received recommendations by multiplying each confidence level by the weighting value associated with the producer module that produced the recommendation.

25. The computer readable medium of claim 23, the executable instructions further comprising, in response to receiving information from the requester system related to the plurality of survived recommendations transmitted to the requester system, modifying each of the weighting values based on the information received from the requester system.

26. The computer readable medium of claim 25, wherein the information received from the requester system includes a plurality of user reaction values, wherein each of the plurality of user reaction values is associated with a different one of the plurality of survived recommendations transmitted to the requester system.

27. The computer readable medium of claim 26, wherein each user reaction value is indicative of a positive user reaction to the recommendation to which the user reaction value is associated.