SYSTEMS AND METHODS FOR CLASSIFYING CONTENT USING MATRIX FACTORIZATION

• US 20090132901A1
• Filed: 10/10/2007
• Published: 05/21/2009
• Est. Priority Date: 11/29/2006
• Status: Active Grant

• Current Assignee: NEC Corporation
• Original Assignee: NEC Laboratories America Inc (NEC Corporation)
• Inventors: Gong, Yihong, Zhu, Shenghuo, Yu, Kai, Chi, Yun
• Granted Patent: US 7,788,264 B2
• Time in Patent Office: Days
• US Class Current: 715/206
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

Systems and methods for classifying documents each having zero or more links thereto include generating a link matrix; generating a document term matrix; and jointly factorizing the document term matrix and the link matrix.

Claims

1. A method for classifying documents each having zero or more links thereto, comprising:

a. generating a link matrix;

b. generating a document term matrix; and

c. jointly factorizing the document term matrix and the link matrix to classify the documents.

2. The method of claim 1, comprising generating one or more indicative factors to represent the documents.

3. The method of claim 1, wherein the document comprises one of:

a Web page, an email and a referenced document.

4. The method of claim 1, comprising using latent semantic space to represent the documents.

5. The method of claim 4, comprising mapping the documents to the latent semantic space by determining link structure and content of the documents.

6. The method of claim 1, comprising representing documents as vectors.

7. The method of claim 6, comprising classifying the vectors.

8. The method of claim 7, comprising generating one or more classification results.

9. The method of claim 6, comprising clustering the vectors.

10. The method of claim 8, comprising generating one or more clustering results.

11. The method of claim 1, comprising generating a document word co-occurrence matrix from the documents.

12. The method of claim 1, wherein the link matrix is generated by matrix factorization to represent web pages as one or more principal factors of the link structure.

13. The method of claim 1, wherein the link matrix is generated by matrix factorization to derive a feature representation Z of web pages based on analyzing a link matrix A, where Z is an n×

1 matrix, with each row being a 1-dimensional feature vector of a web page.

14. The method of claim 13, comprising applying a factorization defined as:

15. The method of claim 14, comprising determining:

16. The method of claim 1, wherein generating a document term matrix comprises applying a bag-of-words approach.

17. The method of claim 1, comprising performing supervised matrix factorization.

18. The method of claim 17, comprising using data labels to guide a matrix factorization.

19. The method of claim 17, comprising determining:

20. The method of claim 19, comprising applying a gradient method to minimize J_s.

21. The method of claim 1, comprising generating a response to a search query based on the jointly factorized document term and link matrices.