Selective sampling of user state based on expected utility

• US 7,979,252 B2
• Filed: 06/21/2007
• Issued: 07/12/2011
• Est. Priority Date: 06/21/2007
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Kapoor, Ashish, Horvitz, Eric J.
• Primary Examiner(s): Proctor; Jason
• Assistant Examiner(s): GUILL, RUSSELL L
• Application Number:
  US11/766,547
• Publication Number:
  US 20080319727A1
• Time in Patent Office: 1,482 Days
• Field of Search: 703/6
• US Class Current: 703/6
• CPC Class Codes: G06F 9/451 Execution arrangements for user interfaces

Abstract

Model enhancement architecture that provides selective sampling of data to enhance model performance where model testing is deemed to be poor. Sampling can include direct interaction with the user while the user is logged-in to the computing system. The system can be used to infer a computer user'"'"'s current interruptability based on computer activity and relevant contextual information. Personalized models can then be created that are utilized to determine a cost of interruption and an expected utility. A modeling component is provided that builds and runs models based on data. The data can be any type of data such as application data, user profile data, tracking data, user state data, user situation data, and so on. A sampling component samples the data based on failure analysis of the model. The architecture is a utility-centric approach to gathering data to maximally enhance the current model.

Claims

1. A computer-implemented system that facilitates model enhancement, comprising:

at least one processor configured to provide a modeling component that builds and runs a model based on data associated with a user, the model indicating interruptability of the user; and

a sampling component that determines a time at which to obtain additional data associated with the user for building the model, the data being obtained by probing the user, and the time being determined based on;

failure analysis of the model; and

a state of the user.

2. The system of claim 1, wherein the sampling component selectively samples the data further based on expected utility.

3. The system of claim 1, wherein the data is associated with a user situation.

4. The system of claim 1, wherein the data is associated with a user state.

5. The system of claim 1, wherein the sampling component is employed to selectively sample the data when the failure analysis indicates that the model performs below a defined criterion.

6. The system of claim 1, wherein the sampling component initiates sampling of a new set of data when the model performs below a predetermined criterion.

7. The system of claim 6, wherein the new set of data includes data sampled from disparate data sources.

8. The system of claim 7, further comprising a selection component that selects the data that is sampled from the disparate data sources.

9. The system of claim 1, further comprising an analysis component that analyzes results of the model and determines the data over which the model is performing below predefined criteria.

10. The system of claim 1, wherein the data sampled includes data associated with user interaction of a computing system.

11. The system of claim 1, wherein the data sampled includes data associated with user biometrics.

12. The system of claim 1, further comprising a machine learning and reasoning component that employs a probabilistic and/or statistical-based analysis to prognose or infer an action that a user desires to be automatically performed.

13. A computer-implemented method of performing selective sampling of data to enhance model performance, comprising:

with at least one processor, creating a model of a state of a user based on a set of data;

testing the model against the set of data to determine performance results; and

initiating selective sampling to obtain additional data by probing the user when the performance results generated from the set of data are below a criterion, the selective sampling occurring at a time determined based on a state of the user.

14. The method of claim 13, wherein selectively sampling performs direct sampling from the user through one or more prompts requesting the user to classify the user'"'"'s state of interruptability.

15. The method of claim 13, wherein selectively sampling is performed over disparate data sources of logged information.

16. The method of claim 13, wherein the state of the user is determined based on an invasiveness factor.

17. The method of claim 13, wherein selectively sampling is performed further based on an expected utility.

18. The method of claim 13, further comprising analyzing the set of data to determine a general category of interest of the set of data that forms a basis for the act of initiating selective sampling of the additional data.

19. The method of claim 13, further comprising learning when to perform the act of initiating based on logged data.

20. A computer-executable system that facilitates model enhancement, comprising:

means, comprising at least one processor, for creating a model of interruptability of a user based on a set of data;

means for testing the model against the set of data to determine performance results;

means, comprising a portion of a lifelong learning framework, for initiating selective sampling of additional data by probing the user when;

the performance results generated from the set of data are below a criterion, and

an expected improvement in the performance results from the model based on additional data obtained by the probing exceeds an expected cost of interrupting the user with the probing; and

means for generating a cost of interruption based on the additional data.