MULTI-TIERED APPROACH TO E-MAIL PRIORITIZATION

• US 20130212047A1
• Filed: 06/15/2012
• Published: 08/15/2013
• Est. Priority Date: 02/10/2012
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Lai, Jennifer C., Lu, Jie, Pan, Shimei, Wen, Zhen
• Granted Patent: US 9,152,953 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G06N 5/00 Computing arrangements using knowledge-based models
• CPC Class Codes: G06Q 10/107 Computer-aided management of electronic mailing [e-mailing]

Abstract

An apparatus for automating a prioritization of an incoming message, including a batch learning module that generates a global classifier based on training data that is input to the batch learning module. A feedback learning module that generates a user-specific classifier based on a plurality of feedback instances. A feature extraction module that receives the incoming message and a topic-based user model, infers a topic of the incoming message based on the topic-based user model, and computes a plurality of contextual features of the incoming message. A classification module that dynamically determines a priority classification strategy for assigning a priority level to the incoming message based on the plurality of contextual features of the incoming message and a weighted combination of the global classifier and the user-specific classifier, and classifies the incoming message based on the priority classification strategy.

Claims

1. An apparatus, comprising:

an input to receive an incoming message;

at least one processor; and

a memory tangibly embodying a set of instructions for automating a prioritization of the incoming message, the instructions comprising;

a batch learning module that generates a global classifier based on training data that is input to the batch learning module;

a feedback learning module that generates a user-specific classifier based on a plurality of feedback instances;

a feature extraction module that receives the incoming message and a topic-based user model, infers a topic of the incoming message based on the topic-based user model, and computes a plurality of contextual features of the incoming message; and

a classification module that dynamically determines a priority classification strategy for assigning a priority level to the incoming message based on the plurality of contextual features of the incoming message and a weighted combination of the global classifier and the user specific classifier, and classifies the incoming message based on the priority classification strategy.

2. The apparatus of claim 1, wherein the classification module determines the priority classification strategy from a plurality of priority classification strategies.

3. The apparatus of claim 2, wherein the plurality of priority classification strategies comprises a dynamic linear combination scheme with instance matching that, based on instance matching between of the contextual features of the incoming message and contextual features of the plurality of feedback instances, assesses a quality of the user-specific classifier and then, based on the assessed quality of the user-specific classifier, assigns a weight to each of the global classifier and the user-specific classifier for a linear combination thereof.

4. The apparatus of claim 2, wherein the plurality of priority classification strategies comprises a dynamic linear combination scheme with instance matching, and

wherein, when the incoming message and a feedback instance of the plurality of feedback instances have at least one of a same sender and subject, the dynamic linear combination scheme with instance matching assigns a same priority to the incoming message as a priority assigned to the feedback instance having at least one of the same sender and subject.

5. The apparatus of claim 4, wherein, when the incoming message does not have at least one of the same sender and subject as any of the plurality of feedback instances, the dynamic linear combination scheme with instance matching assigns a weight to each of the global classifier and the user-specific classifier for a linear combination thereof.

6. The apparatus of claim 1, wherein the plurality of contextual features comprises a people-centric contextual feature and a message-centric contextual feature.

7. The apparatus of claim 1, wherein the user-specific classifier is only updated based on explicit feedback.

8. The apparatus of claim 1, wherein the user-specific classifier is updated based on explicit feedback and implicit feedback.

9. The apparatus of claim 1, wherein the plurality of feedback instances includes explicit feedback instances, each explicit feedback instance indicating a priority ranking of a different message having been received by the input.

10. The apparatus of claim 1, wherein the feature extraction module computes a contextual feature of the plurality of contextual features of the incoming message based on the topic of the incoming message having been inferred by the feature extraction module.

11. The apparatus of claim 1, wherein the feature extraction module computes a contextual feature of the plurality of contextual features of the incoming message based on a combination of a plurality of factors including:

a percentage of messages previously received by the input that have a substantially similar topic as the topic of the incoming message inferred by the feature extraction module;

a percentage of the messages previously received having the substantially similar topic which are determined to have been read; and

a percentage of the messages previously received having the substantially similar topic which are determined to have been at least one of forwarded, replied, saved, and flagged.

12. The apparatus of claim 1, wherein the feature extraction modules computes the plurality of contextual features of the incoming message based on the topic of the incoming message having been inferred by the feature extraction module.

13. The apparatus of claim 1, wherein the plurality of contextual features comprises a message centric feature that is computed based on a percentage of received messages comprising a substantially similar topic as the topic of the incoming message having been inferred by the feature extraction module.

14. A computer system comprising a memory tangibly embodying a set of instructions for automating a prioritization of an incoming message, the instructions causing the computer system to comprise:

a plurality of classifiers comprising;

a global classifier that is created with training data; and

a user-specific classifier that is dynamically updated based on a feedback instance;

a topic-based user model comprising a plurality of topic models;

a feature extraction module that infers a topic of the incoming message and computes a plurality of contextual features of the incoming message based on the inferred topic of the incoming message; and

a classification module that assigns a weight to each contextual feature of the plurality of contextual features based on a dynamic combination of the global classifier and the user specific classifier, combines the assigned weight of each contextual feature, and determines a priority level of the incoming message.

15. The computer system according to claim 14, wherein the classification module determines the priority level based on the combined assigned weight of the contextual features.

16. The computer system according to claim 14, wherein the feature extraction module infers the topic of the incoming message based on one or more topic models of the plurality of topic models.

17. The computer system according to claim 14, wherein the priority level of the incoming message comprises a binary priority level.

18. The computer system according to claim 14, wherein the computer system comprises a first computer and a second computer, and the priority level determined by the classification module is transmitted from the first computer to the second computer.

19. The computer system according to claim 18, wherein the first computer receives the incoming message before the second computer, and when the second computer receives the incoming message, the incoming message has the priority level assigned thereto.

20. A computer system for automating a prioritization of an incoming message, the computer system comprising:

a plurality of classifiers comprising;

a global classifier that is created with training data; and

a user-specific classifier that is dynamically updated based on a feedback instance;

a topic-based user model comprising a plurality of topic models;

a feature extraction module that infers a topic of the incoming message and computes a plurality of contextual features of the incoming message based on the inferred topic of the incoming message; and

a classification module that assigns a weight to a set of contextual features of the plurality of contextual features based on a dynamic combination of the global classifier and the user specific classifier, combines the assigned weight of the set of contextual features, and determines a priority level of the incoming message.

21. The computer system according to claim 20, wherein the classification module determines the priority level based on the combined assigned weight of the set of contextual features.

22. The computer system according to claim 20, wherein the feature extraction module infers the topic of the incoming message based on one or more topic models of the plurality of topic models.

23. The computer system according to claim 20, wherein the priority level of the incoming message comprises a binary priority level.

24. An apparatus, comprising:

an input to receive an incoming message;

at least one processor; and

a memory tangibly embodying a set of instructions for automating a prioritization of the incoming message, the instructions causing the apparatus to comprise;

a feature extraction module that infers a topic of the incoming message based on a topic-based user model, and computes a plurality of contextual features of the incoming message based on the inferred topic of the incoming message; and

a classification module that assigns a weight to the plurality of contextual features based on a dynamic combination of a plurality of classifiers and dynamically determines a priority classification strategy for assigning a priority level to the incoming message based on a combination of the assigned weight of the contextual features.

25. The apparatus according to claim 24, wherein the plurality of classifiers comprises a user-specific classifier and a global classifier,

wherein the apparatus dynamically assesses a quality of the user-specific classifier based on a matching between the incoming message and a plurality of feedback instances, and

wherein, when the matching is above a predetermined threshold, the user-specific classifier is assigned a weight that is greater than a weight assigned to the global classifier.