Systems and methods for estimating and integrating measures of human cognitive load into the behavior of computational applications and services

• US 7,499,896 B2
• Filed: 08/08/2006
• Issued: 03/03/2009
• Est. Priority Date: 04/20/1999
• Status: Expired due to Fees

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Horvitz, Eric J.
• Primary Examiner(s): STARKS, WILBERT L
• Application Number:
  US11/463,198
• Publication Number:
  US 20060294036A1
• Time in Patent Office: 938 Days
• Field of Search: 706/45, 709/206
• US Class Current: 706/45
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/10 using kernel methods, e.g. support vector machines [SVM]
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

The present invention relates to a system and methodology for automatically performing computerized actions. By comparing a received textual input against a list of words in a text classification system an action probability can be determined for automated performance of an action or event, the nonperformance of an action or event, or generating a prompt for approval of an action or event as indicated by the received textual input. Thus, the disclosed subject matter facilitates automated computerized action based on the probability that a user would desire the particular action.

Claims

1. A system for automatically performing computerized actions, comprising:

a text interpretation component that analyzes a received text input for at least one word that indicates a wish to perform an action based in part on action-related content, the analysis is based at least in part on a comparison to at least one word in a text classification data store;

an action probability component that computes a probability value of the action based at least in part upon the comparison of the text interpretation component; and

a decision-based execution component that provides at least one of the following;

an execution of an action based at least in part upon the probability value of the action probability component and a probability threshold;

no execution of an action based at least in part upon the probability value of the action probability component and a probability threshold; and

a prompt to execute an action based at least in part on the probability value of the action probability component and a probability threshold.

2. The system of claim 1, the probability threshold is manually configurable.

3. The system of claim 1, the probability threshold is automatically configured in accordance with a decision theory.

4. The system of claim 1, the text input is generated from a text output of a disparate system.

5. The system of claim 4, further comprises a training component that classifies the text input as at least one of indicating that an action is to occur or indicating that an action is not to occur based at least in part upon evaluating a previous action observed in conjunction with a similar input.

6. The system of claim 5, the training component stores the classification as a case and the action probability component utilizes previous relevant case information in computing the probability value.

7. The system of claim 1, further comprises a delay component that calculates a time delay for the decision-based component to wait before at least one of executing the action or displaying a prompt to execute the action.

8. The system of claim 7, the time delay is determined by a statistical regression based at least in part upon the size of the text input.

9. The system of claim 7, the calculated time is based at least in part upon one of system-generated observations for a given user, a saved profile comprised of such observations with regard to timing preferences, or a manually input timing preference configuration.

10. The system of claim 1, the text input based at least in part on an electronic mail message, the words of the text input are compared to words in the text classification data store that indicate a wish to perform a scheduling action.

11. A method for automatically determining a computerized action to be executed comprising:

comparing action-related content of a received textual input against at least one list of words in a text classification system;

calculating an action probability of an action based at least in part upon the comparison, the action is related to an action of the action-related content;

analyzing the action probability with a probability threshold to determine the action status, the action status is at least one of an inaction, an action with approval, or an automatic action; and

executing the action based at least in part upon the action status.

12. The method of claim 11, further comprising delaying the execution of an action based at least in part on the size of the textual input.

13. The method of claim 12, further comprising learning an appropriate delay based at least in part upon a statistical regression analysis of the length of the textual input or previous observations of the delay between display of a similar textual input and a response resultant from the similar input.

14. The method of claim 13, the appropriate delay is based at least in part upon a goal, a need, or an attention discerned based on a user action.

15. The method of claim 14, the discerned user action is based on at least one of system-generated observations for a given user, a saved profile comprised of such observations with regard to timing preferences, or a manual input of a timing preference configuration.

16. The method of claim 11, further comprising training the text classification system to identify at least one of a positive textual input or a negative textual input based at least in part upon a subsequent user action for a previous textual input.

17. A system for determining appropriate computerized actions and inactions comprising:

means for analyzing a message to determine a word that indicates an event should be performed based in part on action-related content of the message;

means for applying a probabilistic calculation to the message to determine a probability of an action thereby creating an action probability, the action is related to the event;

means for comparing the calculation to a threshold to determine at least one of an action or inaction to implement; and

means for implementing at least one of the action and inaction based at least in part upon the comparison of the probabilistic calculation and the threshold.

18. The system of claim 17, the message is an electronic mail message (email).

19. The system of claim 18, further comprising means for determining a delay based at least in part upon a length of time between display of the email or a response generated in connection with the email.

20. The system of claim 19, further comprising means for waiting to perform the action based upon the delay.