Message and user attributes in a message filtering method and system

• US 6,778,941 B1
• Filed: 11/13/2001
• Issued: 08/17/2004
• Est. Priority Date: 11/14/2000
• Status: Expired due to Term

• Current Assignee: Barracuda Networks Incorporated
• Original Assignee: Qualia Computing Inc. (iCAD Incorporated)
• Inventors: Amburn, Philip, Kabrisky, Matthew, Rogers, Steven K., Worrell, Steven W.
• Primary Examiner(s): Bhat, Aditya S.
• Application Number:
  US10/010,789
• Time in Patent Office: 1,008 Days
• Field of Search: 702/176, 709/206, 709/207, 709/230, 709/202, 715/500, 715/501.1, 707/100
• US Class Current: 702/176
• CPC Class Codes: G06Q 10/107 Computer-aided management of electronic mailing [e-mailing]
• CPC Class Codes: H04L 51/212 using filtering or selective blocking
• CPC Class Codes: H04L 67/30 Profiles
• CPC Class Codes: H04L 69/329 in the application layer [OSI layer 7]
• CPC Class Codes: H04L 9/40 Network security protocols

Abstract

A method and system for filtering messages where the importance of a message is determined by analyzing the message body in conjunction with message attributes. Message body refers to the text in the body of the message, whereas message attributes convey information about the message. In another embodiment, analysis of the user'"'"'s current computing environment provides additional input to the filtering system. This allows for preferentially weighting messages of user'"'"'s current interests. Analysis includes computation of feature vectors and subsequent input to a discriminant function. The discriminant function provides a test statistic which is compared to a threshold. If the test statistic exceeds the threshold, the incoming message is passed by the filtering system and may be displayed to the user. In another embodiment, message body and attributes are used to anticipate significant events in a time series, such as streaming financial data.

Claims

1. A method for message filtering comprising the steps of:

extracting message body data from a message body portion of a message;

extracting message attribute data from the message, where the message attribute data is derived from the group comprising;

message source, author, date, day of week, time of day, corporate affiliation, and academic affiliation;

computing a message feature vector jointly from the message body data and the message attribute data;

computing a message discriminant score using the message feature vector; and

passing or withholding the message based on the discriminant score.

2. The method of claim 1, wherein a prompt is given to the user to indicate when a message is passed.

3. The method of claim 2, wherein the prompt is an audio cue.

4. The method of claim 2, wherein the prompt is a visual cue.

5. The method of claim 2, wherein the prompt is an audiovisual cue.

6. The method of claim 1, wherein the-message is displayed to the user when passing.

7. A method for message filtering comprising the steps of:

extracting message body data from a message body portion of a message;

extracting message attribute data from the message;

computing a message feature vector jointly from the message body data and the message attribute data;

computing user textual features from the user environment;

computing user attribute features from the user environment;

computing a user feature vector jointly from the user textual features and the user attribute features;

computing a message-user similarity score from the message feature vector and the user feature vector;

passing or withholding the message based on the message-user similarity scores; and

wherein the user environment comprises documents currently in use and recently used documents.

8. The method of claim 7, wherein the message-user similarity score is computed according to the formula:

$z_{m,u}=\frac{x_m·<br><br>x_u}{<br><br>x_m<br><br>·<br><br><br><br>x_u<br><br>}.$

9. A method for filtering messages arriving in an online system, the method comprising the steps of:

providing a plurality of incoming messages from an online system to a user;

receiving an input from the user instructing the online system to act upon an incoming message;

labeling each incoming message in response to the instruction from the user to act upon the incoming message to create an online labeled data set;

training a classifier with the online labeled data set; and

wherein the classifier is retrained at predetermined intervals with current online data sets, formed from recently received incoming messages, to provide a classifier for identifying messages of current interest to the user.

10. The method of claim 9, wherein the input from the user comprises an instruction to ignore the incoming message.

11. The method of claim 9, wherein the input from the user comprises an instruction to read and then delete the incoming message.

12. The method of claim 9, wherein the input from the user comprises an instruction to read and perform a further action upon the incoming message.

13. The method of claim 12, wherein the further action comprises forwarding the message.

14. The method of claim 9, wherein the incoming messages are passed to the user by the classifier, and the classifier is retrained using the online labeled data set.

15. The method of claim 9, wherein the step of training a classifier comprises the steps of:

computing feature vectors from messages;

computing feature vectors from the user environment;

computing a preferentially weighted message feature vector according to the formula;