Method and system using keyword vectors and associated metrics for learning and prediction of user correlation of targeted content messages in a mobile environment

• US 20090125462A1
• Filed: 11/11/2008
• Published: 05/14/2009
• Est. Priority Date: 11/14/2007
• Status: Active Grant

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: Verma, Nakul, KRISHNASWAMY, Dilip, Bychkovsky, Vladimir
• Granted Patent: US 9,705,998 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G06F 40/10 Text processing natural language analysis G06F40/20; semantic analysis G06F40/30; processing or translation of natural language G06F40/40
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/004 Artificial life, i.e. computing arrangements simulating life
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: H04L 12/1859 adapted to provide push services, e.g. data channels
• CPC Class Codes: H04L 12/189 in combination with wireless systems selective distribution or broadcast in wireless communication networks H04W4/06
• CPC Class Codes: H04L 51/18 Commands or executable codes
• CPC Class Codes: H04L 51/212 using filtering or selective blocking
• CPC Class Codes: H04L 51/214 using selective forwarding
• CPC Class Codes: H04L 51/58 Message adaptation for wireless communication
• CPC Class Codes: H04L 67/04 specially adapted for terminals or networks with limited capabilities; specially adapted for terminal portability
• CPC Class Codes: H04L 67/306 User profiles
• CPC Class Codes: H04L 67/52 specially adapted for the location of the user terminal
• CPC Class Codes: H04L 67/53 using third party service providers
• CPC Class Codes: H04L 67/535 Tracking the activity of the user network monitoring arrangements H04L43/00; recording of computer activity G06F11/34
• CPC Class Codes: H04L 67/55 Push-based network services
• CPC Class Codes: H04L 67/61 taking into account QoS or priority requirements
• CPC Class Codes: H04L 69/08 Protocols for interworking; Protocol conversion
• CPC Class Codes: H04M 3/4878 Advertisement messages
• CPC Class Codes: H04W 4/12 Messaging; Mailboxes; Announcements
• CPC Class Codes: H04W 4/21 : for social networking applications
• CPC Class Codes: H04W 8/18 : Processing of user or subscriber data, e.g. subscribed services, user preferences or user profiles; Transfer of user or subscriber data

Abstract

Methods and systems for determining a suitability for a mobile client to display information are disclosed. A particular exemplary method includes receiving a plurality of sets of one or more first keywords on a mobile client, each set of first keywords associated with one or more respective first messages, monitoring user interaction of the respective first messages on the mobile client, performing learning operations on the mobile client with the first keywords based on monitored user interaction to estimate a set of keyword interest weights, receiving a set of target keywords associated with a target message, and displaying the target message on the mobile client based on the estimated keyword interest weights.

Claims

1. A method for determining display information on a mobile client, comprising:

receiving a plurality of sets of one or more first keywords on the mobile client, each set of first keywords associated with one or more respective first messages;

monitoring user interaction of the respective first messages on the mobile client;

performing learning operations on the mobile client with the first keywords based on monitored user interaction to estimate a set of keyword interest weights;

receiving a set of target keywords associated with a target message; and

displaying the target message on the mobile client based on the estimated keyword interest weights.

2. The method according to claim 1, further comprising:

performing a prediction routine based on the target keywords and the keyword interest weights to determine an estimated user interest of the target message,

wherein the target message is displayed when the estimated user interest is favorable compared to other estimated user interests of other messages.

3. The method according to claim 2, wherein the step of performing a prediction routine includes performing a correlation operation R={circumflex over (P)}·

A, where {circumflex over (P)} is a current estimate of user interest weights and A is a vector representation of the target message.

4. The method according to claim 1, further comprising receiving the target message over a wireless link if the estimated keyword interest weights indicate a desirability of the target message.

5. The method according to claim 1, wherein the step of monitoring user interaction includes monitoring click-through-rates for the first messages.

6. The method according to claim 1, wherein performing learning operations includes using at least one steepest descent algorithm to estimate at least one keyword interest weights.

7. The method according to claim 1, wherein performing learning operations includes using at least one Newtonian algorithm to estimate at least one keyword interest weights.

8. The method according to claim 1, wherein the keywords are specified from a keyword dictionary.

9. The method according to claim 8, wherein the rate of learning is determined based on at least one of information-space sparseness, rate of presentation of information, value of initial seed, and aspects of a user profile.

10. The method according to claim 8, wherein the keywords are specified from a specified hierarchical keyword dictionary, wherein the hierarchy has two or more levels.

11. The method according to claim 8, wherein the keywords are specified from a specified flat keyword dictionary.

12. The method according to claim 8, wherein the keywords are not associated with the semantics of a specific language.

13. The method according to claim 8, wherein the step of learning incorporates random or pseudo-random noise.

14. The method according to claim 8, wherein a cardinality of the set of target keywords is sparse relative to the size of the keyword dictionary.

15. The method according to claim 1, wherein learning parameters used for the step of learning operations are incorporated into at least one received message.

16. The method according to claim 15, wherein learning parameters include at least one decay constant.

17. The method according to claim 1, wherein only a subset of the set of keywords is used to display the target message on the mobile client.

18. The method according to claim 17, wherein the subset of keywords is determined based upon a threshold interest value.

19. The method according to claim 17, wherein the subset of keywords is determined by using those keywords having the respective top keyword interest weights.

20. A mobile client configured to display information, comprising:

means for receiving a plurality of sets of one or more first keywords on a mobile client, each set of first keywords associated with one or more respective first messages, the receiving circuit further configured to receive a set of target keywords associated with a target message;

means for monitoring user interaction of the respective first messages on the mobile client;

means for performing learning operations on the mobile client with the first keywords based on monitored user interaction to estimate a set of keyword interest weights; and

means for displaying the target message on the mobile client based on the estimated keyword interest weights.

21. The mobile client according to claim 20, further comprising:

means for performing a prediction routine based on the target keywords and the keyword interest weights to determine an estimated user interest of the target message,

wherein the target message is displayed when the estimated user interest is favorable compared to other estimated user interests of other messages.

22. The mobile client according to claim 21, wherein the means for performing a prediction routine by performing at least a correlation operation R={circumflex over (P)}·

A, where {circumflex over (P)} is a current estimate of user interest weights and A is a vector representation of the target message.

23. The mobile client according to claim 20, wherein the means for performing at least one steepest descent algorithm to estimate at least one keyword interest weights.

24. The mobile client according to claim 20, wherein the means for performing at least one Newtonian algorithm to estimate at least one keyword interest weights.

25. The mobile client according to claim 20, wherein the means for performing incorporates random or pseudo-random noise in its learning operations.

26. The mobile client according to claim 20, wherein a cardinality of the set of target keywords is sparse relative to the size of the keyword dictionary.

27. The mobile client according to claim 20, wherein learning parameters used for the step of learning operations are incorporated into at least one received message.

28. The mobile client according to claim 26, wherein learning parameters include at least one decay constant.

29. The mobile client according to claim 20, wherein only a subset of the set of keywords is used to display the target message on the mobile client.

30. The mobile client according to claim 28, wherein the subset of keywords is determined based upon a threshold interest value.

31. The method according to claim 29, wherein the subset of keywords is determined by using those keywords having the respective top keyword interest weights.

33. The mobile client according to claim 20, further comprising:

means for performing a prediction routine based on the target keywords and the keyword interest weights to determine an estimated user interest of the target message;

wherein the target message is displayed when its estimated user interest is favorable compared to other estimated user interests of other messages.

35. The mobile client according to claim 31, wherein the learning means performs at least one steepest descent algorithm to estimate at least one keyword interest weights.

36. The mobile client according to claim 31, wherein the learning means performs incorporates random or pseudo-random noise in its learning operations.

37. The mobile client according to claim 31, wherein a cardinality of the set of target keywords is sparse relative to the size of the keyword dictionary.

38. The method according to claim 20, wherein only a subset of the set of keywords is used to display the target message on the mobile client.

39. The method according to claim 28, wherein the subset of keywords is determined based upon a threshold interest value.

32. A mobile client configured to display information, comprising:

a receiving circuit configured to receive a plurality of sets of one or more first keywords on a mobile client, each set of first keywords associated with one or more respective first messages, the receiving circuit further configured to receive a set of target keywords associated with a target message;

monitoring means for monitoring user interaction of the respective first messages on the mobile client;

learning means for performing learning operations on the mobile client with the first keywords based on monitored user interaction to estimate a set of keyword interest weights; and

a display configured to display the target message on the mobile client based on the estimated keyword interest weights.

34. The mobile client according to claim 32, wherein the prediction means performs a prediction routine by performing at least a correlation operation R={circumflex over (P)}·

A, where {circumflex over (P)} is a current estimate of user interest weights and A is a vector representation of the target message.

40. A computer program product, comprising:

a computer-readable medium comprising instructions for the following operations;

instructions for receiving a plurality of sets of one or more first keywords on a mobile client, each set of first keywords associated with one or more respective first messages;

instructions for monitoring user interaction of the respective first messages on the mobile client;

instructions for performing learning operations on the mobile client with the first keywords based on monitored user interaction to estimate a set of keyword interest weights;

instructions for receiving a set of target keywords associated with a target message; and

instructions for displaying the target message on the mobile client based on the estimated keyword interest weights.