Method for automatically assigning priorities to documents and messages

• US 8,166,392 B2
• Filed: 05/21/2003
• Issued: 04/24/2012
• Est. Priority Date: 07/30/1999
• Status: Expired due to Fees

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Horvitz, Eric
• Primary Examiner(s): Ries, Laurie
• Assistant Examiner(s): Mills, Frank D
• Application Number:
  US10/442,546
• Publication Number:
  US 20070271504A1
• Time in Patent Office: 3,261 Days
• Field of Search: 715/234, 715/255, 706/15, 706/16, 709/207
• US Class Current: 715/255
• CPC Class Codes: G06Q 10/107 Computer-aided management of electronic mailing [e-mailing]
• CPC Class Codes: Y10S 707/99937 Sorting

Abstract

Methods for prioritizing documents, such as email messages, is disclosed. In one embodiment, a computer-implemented method first receives a document. The method assigns a measure of priority to the document, by employing a text classifier such as a Bayesian classifier or a support-vector machine classifier. The method then outputs the priority. In one embodiment, the method includes alerting the user about a document, such as an email message, based on the expected loss associated with delays expected in reviewing the document as compared to the expected cost of distraction and transmission incurred with alerting the user about the document.

Claims

1. A computer-implemented method comprising:

operating a computer to perform acts comprising;

receiving a first message;

detecting user input indicating one of a plurality of actions with respect to the first message;

training a classifier based on the received first message and the detected user action;

receiving a second message;

generating a priority of the second message with the classifier, and

outputting the priority, wherein outputting the priority comprises;

determining an expected loss of non-review of the second message at a current time;

determining an expected cost of alerting the user of the second message at the current time; and

,

alerting the user of the second message upon determining that the expected loss is greater than the expected cost.

2. The method of claim 1, wherein receiving a message comprises receiving an email.

3. The method of claim 1, wherein the classifier comprises a Bayesian classifier.

4. The method of claim 1, wherein the classifier comprises a support-vector machine classifier.

5. The method of claim 1, wherein training the classifier further comprises training the classifier to discriminate features of a message that are considered important, the training being based on messages detected as having been opened by a user before other messages.

6. The method of claim 1, wherein generating a priority comprises generating a priority as a likelihood the second message is of high priority.

7. The method of claim 1, wherein outputting the priority comprises alerting the user of the second message based on the priority.

8. The method of claim 1, wherein, the one of the plurality of actions comprising at least deleting the first message without opening the first message.

9. The method of claim 1, wherein determining an expected loss of non-review of the second message at a current time considers the expected loss of non-review of the second message at a future time when the user would otherwise review the message.

10. The method of claim 1, wherein determining an expected loss comprises initially determining a current expected rate of lost opportunity as a function of present time.

11. The method of claim 1, wherein determining an expected cost comprises determining a probability inference as to a likelihood of the user being present.

12. The method of claim 1, wherein training the classifier comprises continually training the message classifier.

13. The method of claim 1, wherein the method further comprises maintaining a plurality of training folders, each folder having a semantic label indicative of a message priority, a semantic label on a folder of the plurality of training folders comprises an indication of a user context and time until message delivery in that context.

14. A computer implemented method comprising:

receiving a plurality of messages comprising text;

receiving sensor input, the sensor input indicating a user context;

receiving user input associated with each of at least a portion of the plurality of messages, the user input comprising, for each message, an indication of whether the user wishes to be interrupted with that message in an indicated context;

based on the user input and the text of the portion of the plurality of messages, training a message classifier, the message classifier comprising one of a Bayesian classifier and a support vector machine classifier, to associate a priority with a received message based on text of the message and a context of the user at a time of receipt of the received message;

receiving a subsequent message at a subsequent time;

generating a priority of the subsequent message with the message classifier and based on an indication of the user context at the subsequent time, as determined based at least in part on the sensor input; and

outputting the priority.

15. The method of claim 14, further comprising continually training the message classifier.

16. The method of claim 14, wherein the message classifier comprises seeding the message classifier to discriminate features of a message that are considered important.

17. The method of claim 14, wherein generating a priority comprises generating a priority as a likelihood the subsequent message is of high priority.

18. The method of claim 14, wherein outputting the priority comprises:

determining an expected loss of non-review of the subsequent message at a current time;

determining an expected cost of alerting the user of the subsequent message at the current time; and

,

alerting the user of the subsequent message upon determining that the expected loss is greater than the expected cost.

19. The method of claim 18, wherein determining an expected loss of non-review of the subsequent message at a current time considers the expected loss of non-review of the subsequent message at a future time when the user would otherwise review the subsequent message.

20. The method of claim 18, wherein determining an expected loss comprises initially determining a current expected rate of lost opportunity as a function of present time.

21. The method of claim 18, wherein determining an expected cost comprises determining a probability inference as to a likelihood of the user being present.

22. The method of claim 14, wherein receiving user input comprises receiving user input indicating a training folder into which to store each message of the at least a portion of the plurality of messages.

23. A machine-readable storage device having instructions stored thereon for execution by a processor to perform a method comprising:

receiving sensor input, the sensor input indicating a user context;

receiving a plurality of messages comprising text;

receiving user input associated with each of at least a portion of the plurality of messages, the user input comprising, for each message, an indication of whether the user wishes to be interrupted with that message in an indicated context;

based on the user input and the text of the portion of the plurality of messages, training a message classifier to associate a priority with a received message based on text of the message and a context of the user, as indicated by the sensor input, at a time of receipt of the received message;

receiving a subsequent message at a subsequent time;

generating a priority of the subsequent message based on the message classifier and an indication of user context at the subsequent time; and

outputting the priority.

24. The storage device of claim 23 wherein receiving a subsequent message comprises receiving an email.

25. The storage device of claim 23 wherein the message classifier comprises one of a Bayesian classifier and a support vector machine classifier.

26. The storage device of claim 23 further initially comprising training the message classifier by seeding the-message classifier to discriminate features of a message that are considered important.

27. The storage device of claim 23 wherein outputting the priority comprises:

determining an expected loss of non-review of the subsequent message at a current time;

determining an expected cost of alerting the user of the subsequent message at the current time; and

,

alerting the user of the subsequent message upon determining that the expected loss is greater than the expected cost.

28. The storage device of claim 27, wherein determining an expected loss of non-review of the subsequent message at a current time considers the expected loss of non-review of the subsequent message at a future time when the user would otherwise review the message.

29. The storage device of claim 27, wherein determining an expected loss comprises initially determining a current expected rate of lost opportunity as a function of present time.

30. The storage device of claim 27, wherein determining an expected cost comprises determining a probability inference as to a likelihood of the user being present.