SYSTEM AND METHOD FOR CLASSIFYING COMMUNICATIONS THAT HAVE LOW LEXICAL CONTENT AND/OR HIGH CONTEXTUAL CONTENT INTO GROUPS USING TOPICS

• US 20120166179A1
• Filed: 12/27/2010
• Published: 06/28/2012
• Est. Priority Date: 12/27/2010
• Status: Active Grant

• Current Assignee: Avaya Incorporated
• Original Assignee: Avaya Incorporated
• Inventors: Dhara, Krishna Kishore, Krishnaswamy, Venkatesh, Shim, Eunsoo, Wu, Xiaotao, Tirumalachetty, Vikranth Babu
• Granted Patent: US 8,868,406 B2
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G06F 40/258 Heading extraction; Automatic titling; Numbering

Abstract

Disclosed herein are systems, methods, and non-transitory computer-readable storage media for identifying document topics. A system configured to practice the method receives a document from a corpus of documents, learns interpersonal relationships of users associated with the document, performs a lexical analysis of the document, and, based on the interpersonal relationships of the users and the lexical analysis, identifying a topic for the document. The approaches disclosed herein can integrate user-people relationships to identify topics for documents with low lexical or high contextual content. The system can learn this user-people relationship from context. The system uses this learned behavior to identify communication documents correctly. Another aspect is the separation of the two phases. The system overlays the learned model on the lexical topic analysis, allowing the system to capture user-defined topics and user behavior that is learned from other factors such as medium (calls, events, etc) or user preferences.

Claims

1. A method of identifying document topics, the method comprising:

receiving a document from a corpus of documents;

learning interpersonal relationships of users associated with the document;

performing a lexical analysis of the document; and

based on the interpersonal relationships of the users and the lexical analysis, identifying a topic for the document.

2. The method of claim 1, wherein the document has at least one of low lexical content and high contextual content.

3. The method of claim 1, further comprising generating a matrix based on the lexical analysis and the interpersonal relationships and identifying the topic for the document based on the matrix.

4. The method of claim 3, wherein the matrix comprises a lexical document-topic relationship matrix, a topic-keywords matrix, and a user-topic matrix.

5. The method of claim 4, wherein the lexical document-topic relationship matrix and the topic-keywords matrix are generated using Latent Dirichlet Allocation.

6. The method of claim 4, wherein the user-topic matrix is generated based on a participant list for each document in the corpus and based on the lexical document-topic relationship matrix and the topic-keywords matrix.

7. The method of claim 4, wherein the topic for the document is identified based on a balance of lexical content and interpersonal relationships.

8. A system for classifying documents by topic, the system comprising:

a processor;

a first module configured to control the processor to receive a document from a corpus of documents;

a second module configured to control the processor to learn interpersonal relationships of users associated with the document;

a third module configured to control the processor to perform a lexical analysis of the document; and

a fourth module configured to control the processor to classify the document by topic based on the interpersonal relationships of the users and the lexical analysis.

9. The system of claim 8, wherein the document has at least one of low lexical content and high contextual content.

10. The system of claim 8, further comprising generating a matrix based on the lexical analysis and the interpersonal relationships and identifying the topic for the document based on the matrix.

11. The system of claim 10, wherein the matrix comprises a lexical document-topic relationship matrix, a topic-keywords matrix, and a user-topic matrix.

12. The system of claim 11, wherein the lexical document-topic relationship matrix and the topic-keywords matrix are generated using Latent Dirichlet Allocation.

13. The system of claim 11, wherein the user-topic matrix is generated based on a participant list for each document in the corpus and based on the lexical document-topic relationship matrix and the topic-keywords matrix.

14. The system of claim 11, wherein the topic for the document is identified based on a balance of lexical content and interpersonal relationships.

15. A non-transitory computer-readable storage medium storing instructions which, when executed by a computing device, cause the computing device to identify document topics, the instructions comprising:

receiving a document from a corpus of documents;

learning interpersonal relationships of users associated with the document;

performing a lexical analysis of the document; and

based on the interpersonal relationships of the users and the lexical analysis, identifying a topic for the document.

16. The non-transitory computer-readable storage medium of claim 15, wherein the document has at least one of low lexical content and high contextual content.

17. The non-transitory computer-readable storage medium of claim 15, further comprising generating a matrix based on the lexical analysis and the interpersonal relationships and identifying the topic for the document based on the matrix.

18. The non-transitory computer-readable storage medium of claim 17, wherein the matrix comprises a lexical document-topic relationship matrix, a topic-keywords matrix, and a user-topic matrix.

19. The non-transitory computer-readable storage medium of claim 18, wherein the lexical document-topic relationship matrix and the topic-keywords matrix are generated using Latent Dirichlet Allocation.

20. The non-transitory computer-readable storage medium of claim 18, wherein the user-topic matrix is generated based on a participant list for each document in the corpus and based on the lexical document-topic relationship matrix and the topic-keywords matrix.