SYSTEM AND METHOD FOR RECOMMENDING EDUCATIONAL RESOURCES

• US 20100159438A1
• Filed: 12/19/2008
• Published: 06/24/2010
• Est. Priority Date: 12/19/2008
• Status: Active Grant

• Current Assignee: PARC-Palo Alto Research Center, Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: PARC-Palo Alto Research Center, Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Cifarelli, Claudio, Do, Minh, Price, Robert Roy, Perronnin, Florent, Zhou, Rong, Lofthus, Robert M., German, Kristine A.
• Granted Patent: US 8,457,544 B2
• Time in Patent Office: Days
• US Class Current: 434/433
• CPC Class Codes: G06F 16/24573 using data annotations, e.g. user-defined metadata
• CPC Class Codes: G06F 16/24578 using ranking
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G09B 5/00 Electrically-operated educational appliances working with questions and answers G09B7/00; simulators G09B9/00; advertising or displaying in general G09F

Abstract

A recommender system and method including receiving a request identifying at least one current student for which a recommendation is sought and accessing stored student data including attributes associated with respective students of the current students and a plurality of predecessor students. For the predecessor students the attributes include educational resources. The method includes clustering one of predecessor students and the current students into clusters based on sameness of first selected attributes of their respective associated attributes. The method includes generating a mapping between respective students of the other of the predecessor and current students and one of the clusters based on sameness of second selected attributes associated with the students being mapped. The method then includes recommending for each of the current students the educational resource associated with the predecessor student or cluster of predecessor students that the current student is mapped with.

Claims

1. A recommender system for making recommendations in educating students, the system comprising a processor executing a series of programmable instructions for:

receiving a request identifying at least one current student for which a recommendation is sought;

accessing a data storage facility storing student data including at least one attribute associated with respective students of the at least one current student and a plurality of predecessor students, wherein for respective predecessor students of the plurality of predecessor students the at least one attribute includes at least one educational resource used with the predecessor student;

clustering one of predecessor students of the plurality of predecessor students and the at least one current student into a selected number of clusters based on sameness of at least one first selected attribute of their respective associated at least one attributes, wherein the selected number is at least two;

generating a mapping between respective students of the other of the predecessor and current students and one of the at least two clusters based on sameness of at least one second selected attribute associated with the students being mapped; and

recommending for each of the current students the at least one educational resource associated with the predecessor student or cluster of predecessor students that the current student is mapped with.

2. The recommender system according to claim 1, wherein each student of the at least one current student and the plurality of predecessor students is associated with a location in a multidimensional space, wherein each dimension of the multidimensional space is associated with an attribute, the location is defined by the selected attributes of the student, and sameness of the at least one first selected attributed associated with students being clustered is based on a relationship between the locations associated with them.

3. The recommender system according to claim 2, wherein the at least one selected attribute associated with the students in each cluster of the at least two clusters defines a region within the multidimensional space for that cluster, wherein the sameness of the at least one second selected attributed associated with a student and cluster being mapped is based on a relationship between the location associated with the student and the region defined by cluster.

4. The recommender system according to claim 3, wherein the relationship between the locations for clustering and the relationship between the location and the region for mapping is a degree of closeness.

5. The recommender system according to claim 4, wherein the closeness of a location to a region is a closeness of the location to a location within the region that satisfies a pre-selected condition.

6. The recommender system according to claim 5, wherein the location within the region is associated with a mean of the region.

7. The recommender system according to claim 3, wherein each student and cluster is associated with a similarity matrix and wherein the sameness for mapping between two entities selected from a student and a cluster, and the sameness for clustering two students is determined based on a dot product of the associated similarity matrices.

8. The recommender system according to claim 1, wherein the at least one stored attribute associated with respective predecessor students of the plurality of predecessor students further includes assessment data indicating a level of performance in association with the at least one educational resource used by the predecessor student, and the predecessor students involved in the mapping have assessment data indicating a level of performance that exceeds a pre-selected threshold.

9. The recommender system in accordance with claim 8, wherein the clustering is performed only on the predecessor students having assessment data indicating a level of performance that exceeds a pre-selected threshold.

10. The recommender system according to claim 8, wherein the processor further executes programmable instructions for ranking the at least one educational resource recommended for each current student based on the level of performance associated with the educational resource for each predecessor student that the educational resource is associated with.

11. The recommender system according to claim 8, wherein at least one constraint applied to the clustering or mapping is user selectable, and wherein the at least one constraint is selected from the group of constraints consisting of:

the relationship used for determining sameness between students being clustered, the relationship for determining the sameness in mapping, the pre-selected condition, the selected number of clusters, the at least one first selected attribute, the at least one second selected attribute, the pre-selected threshold for the level of performance, the degree of closeness, a maximum size of each cluster, a minimum size of each cluster, an algorithm to use for the clustering, and an algorithm to use for the mapping, which of the at least one current student or plurality of predecessor students is clustered or mapped, a degree of differentiation in sameness between clusters, and criteria for ranking the educational resources.

12. The recommender system in accordance with claim 11,

wherein the at least one constraint is specified in the request;

wherein the processor further executes programmable instructions for;

outputting the result of the recommending to a user;

receiving adjustments to the constraints; and

repeating the clustering, determining, correlating, recommending, and outputting until the user is satisfied with the output results.

13. The recommender system in accordance with claim 12, wherein each of the at least one constraints has an associated weight or set of weights which is applied to the constraint during the clustering or mapping, and wherein the receiving adjustments includes receiving adjustments to the weights associated with at least one of the at least one constraint.

14. The recommender system in accordance with claim 1, wherein the at least one attribute associated with the predecessor students of the plurality of predecessor students and the at least one attribute associated with the current students of the at least one current student includes granular assessment data obtained from markings extracted from assessment answer sheets scanned in via a multifunctional device (MFD), wherein the assessment answer sheets are used for recording answers to questions associated with an assessment of the student.

15. The recommender system in accordance with claim 1, wherein the clustering includes using sparsification.

16. The recommender system in accordance with claim 1, wherein the clustering includes applying logistical constraints which must be satisfied by the clustering.

17. The recommender system in accordance with claim 1, wherein the clustering includes statistical clustering.

18. The recommender system in accordance with claim 1, wherein the clustering includes maximizing a difference between the at least one attribute of the clusters.

19. The recommender system in accordance with claim 21, wherein each cluster has a mean, wherein the difference is maximized by maximizing the difference in the at least one attribute between mean of the respective clusters.

20. The recommender system according to claim 1, wherein the plurality of predecessor students includes a relatively large number of students and the at least one current student includes a relatively small number of students.

21. The recommender system in accordance with claim 1, wherein the clustering includes using non-negative matrix factorization.

22. The recommender system in accordance with claim 1, wherein the clustering includes using probabilistic latent semantic analysis.

23. The recommender system in accordance with claim 1,

wherein each student of the plurality of students is associated with a D-dimensional vector encoding granular assessment data for that student, wherein the vector includes values indicative of an evaluation of each answer to D respective questions, wherein the value for each question of the D questions is +1 for a correct answer and −

1 for an incorrect answer; and

wherein the clustering includes determining the sameness of a first student having an associated vector u and a second student having an associated vector v, and applying a formula to U and v.

24. The recommender system in accordance with claim 23, wherein the formula is:

0.5*(1+cos(u,v)).

25. A recommender system for making recommendations in educating students, the system comprising a processor executing a series of programmable instructions for:

receiving a request identifying at least two current students for which a recommendation is sought, indicating at least two educational resources, and requesting a recommendation for using the at least two educational resources with the at least two current students;

accessing a data storage facility storing educational resource data including at least one attribute associated with respective educational resources of a plurality of educational resources, and student data including at least one attribute associated with respective students of the at least one current student and a plurality of predecessor students, wherein for the plurality of predecessor students the at least one attribute indicates at least one educational resource used by the respective predecessor student and assessment data indicating a level of performance in association with the at least one educational resource used by the predecessor student;

selecting predecessor students of the plurality of predecessor students having assessment data indicating a level of performance that exceeds a pre-selected threshold in association with the indicated at least two educational resources;

clustering one of the selected predecessor students and the at least one current student into a selected number of clusters based on sameness of at least one first selected attribute of their respective associated at least one attributes, wherein the selected number is at least two;

generating a mapping between respective students of the other of the selected predecessor and the at least one current student and one of the at least two clusters based on sameness of at least one second selected attribute associated with the students and clusters being mapped; and

recommending for each of the current students the at least one educational resource associated with the predecessor student or cluster of predecessor students that the current student is mapped with.

26. The recommender system according to claim 25, wherein the at least two educational resources indicated in the request are selected from the group of educational resources consisting of educators, physical educational material, digital educational material, and instructional methods.

27. The recommender system according to claim 25, wherein the indicating the at least two educational resources in the request includes identifying the at least two educational resources.

28. The recommender system according to claim 25, wherein the indicating the at least two educational resources in the request includes providing at least one attribute associated with each of the at least two indicated educational resources, and wherein the selecting includes determining at least one educational resource having attributes that have at least a pre-selected degree of similarity with the at least one attribute provided in the request, and selecting predecessor students of the plurality of predecessor students having assessment data indicating a level of performance that exceeds a pre-selected threshold in association with the determined at least one educational resources.

29. The recommender system according to claim 25, wherein the selecting includes determining at least one educational resource having attributes that have at least a pre-selected degree of similarity with attributes of the at least two educational resources indicated in the request, and selecting predecessor students of the plurality of predecessor students having assessment data indicating a level of performance that exceeds a pre-selected threshold in association with the determined at least one educational resources.

30. A method for making recommendations in educating students, the method comprising:

receiving a request identifying at least one current student for which a recommendation is sought;

accessing a computerized data storage facility storing student data including at least one attribute associated with respective students of the at least one current student and a plurality of predecessor students, wherein for the plurality of predecessor students the at least one attribute includes at least one educational resource;

clustering in a computer process one of predecessor students of the plurality of predecessor students and the at least one current student into a selected number of clusters based on sameness of at least one first selected attribute of their respective associated at least one attributes, wherein the selected number is at least two;

generating in a computer process a mapping between respective students of the other of the predecessor and current students and one of the at least two clusters based on sameness of at least one second selected attribute associated with the students being mapped;

recommending for each of the current students the at least one educational resource associated with the predecessor student or cluster of predecessor students that the current student is mapped with.

31. The method according to claim 30,

wherein the at least one stored attribute associated with respective predecessor students of the plurality of predecessor students further includes assessment data indicating a level of performance in association with the at least one educational resource used by the predecessor student;

the method further comprises selecting predecessor students having assessment data that indicates a level of performance that exceeds a pre-selected threshold; and

only selected predecessor students are involved in the mapping.

32. The method according to claim 31, further comprising ranking the at least one educational resource recommended for each current student based on the level of performance associated with the educational resource for each predecessor student that the educational resource is associated with.

33. The method according to claim 30, wherein the clustering and mapping is performed in accordance with at least one user selected weighted constraint, the method further comprising the steps of:

outputting the result of the recommending to a user;

receiving adjustments to the constraints; and

repeating the clustering, determining, correlating, recommending, and outputting until the user is satisfied with the output results.

34. A computer-readable medium storing a series of programmable instructions configured for performing a method of making recommendations in educating students, the method comprising the steps of:

receiving a request identifying at least one current student for which a recommendation is sought;

accessing a computerized data storage facility storing student data including at least one attribute associated with respective students of the at least one current student and a plurality of predecessor students, wherein for the plurality of predecessor students the at least one attribute includes at least one educational resource;

clustering in a computer process one of predecessor students of the plurality of predecessor students and the at least one current student into a selected number of clusters based on sameness of at least one first selected attribute of their respective associated at least one attributes, wherein the selected number is at least two;

generating in a computer process a mapping between respective students of the other of the predecessor and current students and one of the at least two clusters based on sameness of at least one second selected attribute associated with the students being mapped;

recommending for each of the current students the at least one educational resource associated with the predecessor student or cluster of predecessor students that the current student is mapped with.

35. The computer readable medium according to claim 34,

wherein the at least one stored attribute associated with respective predecessor students of the plurality of predecessor students further includes assessment data indicating success of performance in association with the at least one educational resource used by the predecessor student;

the programmable instructions further comprise the step of selecting predecessor students having assessment data that indicates a level of performance that exceeds a pre-selected threshold; and

only selected predecessor students are involved in the mapping.

36. The computer-readable medium according to claim 34, wherein the programmable instructions further comprise the step of ranking the at least one educational resource recommended for each current student based on the level of performance associated with the educational resource for each predecessor student that the educational resource is associated with.