System for determining degrees of similarity in email message information

• US 20050160148A1
• Filed: 01/16/2004
• Published: 07/21/2005
• Est. Priority Date: 01/16/2004
• Status: Active Grant

• Current Assignee: Gozoom.com, Inc.
• Original Assignee: Mailshell, Inc.
• Inventors: Yu, Tonny
• Granted Patent: US 7,590,694 B2
• Time in Patent Office: Days
• US Class Current: 709/206
• 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

Abstract

Similarity of email message characteristics is used to detect bulk and spam email. A determination of “sameness” for purposes of both bulk and spam classifications can use any number and type of evaluation modules. Each module can include one or more rules, tests, processes, algorithms, or other functionality. For example, one type of module may be a word count of email message text. Another module can use a weighting factor based on groups of multiple words and their perceived meanings. In general, any type of module that performs a similarity analysis can be used. A preferred embodiment of the invention uses statistical analysis, such as Bayesian analysis, to measure the performance of different modules against a known standard, such as human manual matching. Modules that are performing worse than other modules can be valued less than modules having better performance. In this manner, a high degree of reliability can be achieved. To improve performance, if a message is determined to be the same as a previous message, the previous computations and results for that previous message can be re-used. Users can be provided with options to customize or regulate bulk and spam classification and subsequent actions on how to handle the classified email messages.

Claims

1. A method for classifying email messages, the method comprising

using a plurality of modules to determine a level of sameness of a particular email message with one or more prior email messages, wherein the level of sameness is derived for the particular email message from a weighting of the outputs of the modules;

determining a performance level for each of the modules;

comparing performance levels;

adjusting a weighting of at least one module in response to comparing performance levels; and

using the level of sameness for the particular email message to classify the particular email message into a category.

2. The method of claim 1, further comprising

comparing the number of email messages classified in the category with a predetermined number;

and

if the number of email messages is greater than the predetermined number then

classifying the category as a first category type;

else

classifying the category as a second category type.

3. The method of claim 2, wherein the first category type is bulk email.

4. The method of claim 2, further comprising

accepting a signal from a user input device to indicate processing of email messages in a category.

5. The method of claim 4, wherein the processing includes preventing the email messages in a category from being delivered to a user.

7. The method of claim 1, wherein a category is commercial email.

8. The method of claim 1, wherein Bayesian analysis is used.

9. The method of claim 1, further comprising

accepting a signal from a user input device to set a parameter;

and

using the parameter to adjust a weighting.

10. The method of claim 1, wherein a module analyzes word count in an email message.

11. The method of claim 1, wherein a module analyzes similarity of text in an email message.

12. The method of claim 1, wherein a module analyzes a similarity of sender addresses.

13. The method of claim 1, wherein a module analyzes a similarity of network routing.

14. The method of claim 1, wherein a module uses a hash of information in an email message.

15. The method of claim 1, wherein a message classification in a bulk category includes a determination of whether the number of email messages in a category exceed a predetermined number, the method further comprising

submitting email messages in the bulk category to analysis to determine a level of commercial text.

16. The method of claim 15, further comprising

preventing messages with a predetermined level of commercial text from being sent to an intended recipient.

17. The method of claim 15, further comprising

intercepting email messages from being sent to an intended recipient;

collecting the intercepted messages for a period of time;

determining whether the collected messages are bulk messages, and if so, submitting the email messages in the bulk category to analysis to determine a level of commercial text.

18. The method of claim 17, further comprising

preventing messages with a predetermined level of commercial text from being sent to an intended recipient.

19. The method of claim 1, further comprising

assigning a lower rating to a module with a low performance level.

20. The method of claim 1, further comprising

assigning a higher rating to a module with a high performance level.

21. The method of claim 1, further comprising

preventing a module with a low performance level from being used in a subsequent determination of a level of sameness.

22. An apparatus for classifying email messages, the apparatus comprising

a processor for executing instructions included in a machine-readable medium, the machine-readable medium including

one or more instructions for using a plurality of modules to determine a level of sameness of a particular email message with one or more prior email messages, wherein the level of sameness is derived for the particular email message from a weighting of the outputs of the modules;

one or more instructions for determining a performance level for each of the modules;

one or more instructions for comparing performance levels;

one or more instructions for adjusting a weighting of at least one module in response to comparing performance levels; and

one or more instructions for using the level of sameness for the particular email message to classify the particular email message into a category.

23. A machine-readable medium including instructions executable by a processor for classifying email messages, the machine-readable medium including

one or more instructions for using a plurality of modules to determine a level of sameness of a particular email message with one or more prior email messages, wherein the level of sameness is derived for the particular email message from a weighting of the outputs of the modules;

one or more instructions for determining a performance level for each of the modules;

one or more instructions for comparing performance levels;

one or more instructions for adjusting a weighting of at least one module in response to comparing performance levels; and

one or more instructions for using the level of sameness for the particular email message to classify the particular email message into a category.

24. An apparatus for classifying email messages, the apparatus comprising

means for using a plurality of modules to determine a level of sameness of a particular email message with one or more prior email messages, wherein the level of sameness is derived for the particular email message from a weighting of the outputs of the modules;

means for determining a performance level for each of the modules;

means for comparing performance levels;

means for adjusting a weighting of at least one module in response to comparing performance levels; and

means for using the level of sameness for the particular email message to classify the particular email message into a category.