Filtering system for providing personalized information in the absence of negative data

• US 8,060,507 B2
• Filed: 01/07/2011
• Issued: 11/15/2011
• Est. Priority Date: 04/04/2000
• Status: Expired due to Fees

• Current Assignee: Yahoo Assets LLC
• Original Assignee: AOL Inc. (Apollo Global Management, Inc.)
• Inventors: Alspector, Joshua, Kolcz, Aleksander
• Primary Examiner(s): Alam; Shahid
• Application Number:
  US12/987,046
• Publication Number:
  US 20110125578A1
• Time in Patent Office: 312 Days
• Field of Search: 707/727, 707/728, 707/731, 707/769, 707/770, 707/802, 707/803, 709/217, 709/219, 705/14.72, 705/14.73, 705/26.1, 706/45
• US Class Current: 707/731
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06Q 30/0255 based on user history
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access
• CPC Class Codes: Y10S 707/99937 Sorting

Abstract

Systems and methods are provided for personalizing advertising for a user. In accordance with certain implementations, information is accessed indicating which documents were selected by a user and which documents were not selected by a user. At least one positive word vector is generated using words contained in at least one of the selected documents, and at least one negative word vector is generated using words contained in at least one of the unselected documents. Document word vectors are generated, and a document rank order is established based on a vector space relationship analysis. Categories associated with the documents are ranked based on the document rank order, and the ranked categories are sent to an ad server. Advertising material associated with the ranked categories may then be received from the ad server in a selected context.

Claims

1. A method for use in personalizing advertising for a user, the method comprising:

accessing information indicating which documents in a set of documents were selected by a user for viewing and which documents in the set of documents were not selected by the user for viewing;

generating at least one positive word vector using words contained in at least a segment of at least one of the documents that was selected by the user for viewing;

generating at least one negative word vector using words contained in at least a segment of at least one the documents that was not selected by the user for viewing;

generating document word vectors for at least some of the documents that were selected by the user for viewing;

performing, using at least one processor, a vector space relationship analysis of the positive word vector, the negative word vector, and the document word vectors;

establishing, using at least one processor, a document rank order of the documents selected by the user for viewing based on the performed vector space relationship analysis;

classifying the documents selected by the user for viewing in predetermined categories;

ranking the predetermined categories based on the document rank order;

sending the ranked categories to an ad server; and

receiving advertisements associated with the ranked categories from the ad server in a selected context.

2. The method of claim 1, wherein the documents in the set of documents are available at a website.

3. The method of claim 1, wherein the segment is one or more of a headline of the document, a lead of the document, or a summary of the document.

4. The method of claim 1, further comprising:

applying an expectation maximization algorithm to the documents in the document set that were not selected by the user for viewing to generate negative labels for certain ones of the non-selected documents, and positive labels for the rest of the non-selected documents; and

wherein generating at least one negative word vector comprises generating at least one negative word vector using words contained in at least a segment of the documents for which the expectation maximization algorithm generates negative labels.

5. The method of claim 1, further comprising:

accessing a document dictionary that contains dictionary words; and

wherein generating document word vectors comprises, for each document in the group of documents, storing a count of the number of occurrences of each dictionary word found in the document.

6. The method of claim 1, wherein performing a vector space relationship analysis comprises training a support vector machine with the positive word vector and the negative word vector and inputting the document word vectors into the trained support vector machine.

7. The method of claim 1, wherein performing a vector space relationship analysis comprises applying Rocchio'"'"'s method to the positive word vector, the negative word vector, and the document word vectors.

8. The method of claim 1, wherein performing a vector space relationship analysis comprises applying Naï

ve Bayes techniques to the positive word vector, the negative word vector, and the document word vectors.

9. The method of claim 1, further comprising:

training a support vector machine by inputting documents labeled with the predetermined categories; and

wherein classifying the documents selected by the user for viewing comprises using the trained support vector machine to label the documents with the predetermined categories.

10. The method of claim 1, wherein the context is a time.

11. The method of claim 1, wherein the context is a location.

12. An article of manufacture comprising a computer-readable storage device storing computer-readable instructions which, when executed, cause one or more processing devices to perform the following:

access information indicating which documents in a set of documents were selected by a user for viewing and which documents in the set of documents were not selected by the user for viewing;

generate at least one positive word vector using words contained in at least a segment of at least one of the documents that was selected by the user for viewing;

generate at least one negative word vector using words contained in at least a segment of at least one of the documents that was not selected by the user for viewing;

generate document word vectors for at least some of the documents that were selected by the user for viewing;

perform, using at least one processor, a vector space relationship analysis of the positive word vector, the negative word vector, and the document word vectors;

establish, using at least one processor, a document rank order of the documents selected by the user for viewing based on the performed vector space relationship analysis;

classify the documents selected by the user for viewing in predetermined categories;

rank the predetermined categories based on the document rank order;

send the ranked categories to an ad server; and

receive advertisements associated with the ranked categories from the ad server in a selected context.

13. The manufacture of claim 12, wherein the documents in the set of documents are available at a website.

14. The manufacture of claim 12, wherein the segment is one or more of a headline of the document, a lead of the document, or a summary of the document.

15. The manufacture of claim 12, wherein the instructions further comprise instructions that cause the one or more processing devices to:

apply an expectation maximization algorithm to the documents in the document set that were not selected by the user for viewing to generate negative labels for certain ones of the non-selected documents, and positive labels for the rest of the non-selected documents; and

wherein, to generate at least one negative word vector, the instructions include instructions that cause the one or more processing devices to generate at least one negative word vector using words contained in at least a segment of the documents for which the expectation maximization algorithm generates negative labels.

16. The manufacture of claim 12, wherein the instructions further comprise instructions that cause the one or more processing devices to:

access a document dictionary that contains dictionary words; and

wherein, to generate document word vectors, the instructions include instructions that cause the one or more processing devices to, for each document in the group of documents, store a count of the number of occurrences of each dictionary word found in the document.

17. The manufacture of claim 12, wherein, to perform a vector space relationship analysis, the instructions further include instructions that cause the one or more processing devices to train a support vector machine with the positive word vector and the negative word vector and input the document word vectors into the trained support vector machine.

18. The manufacture of claim 12, wherein, to perform a vector space relationship analysis, the instructions further include instructions that cause the one or more processing devices to apply Rocchio'"'"'s method to the positive word vector, the negative word vector, and the document word vectors.

19. The manufacture of claim 12, wherein, to perform a vector space relationship analysis, the instructions further include instructions that cause the one or more processing devices to apply Naï

ve Bayes techniques to the positive word vector, the negative word vector, and the document word vectors.

20. The manufacture of claim 12, wherein the instructions further comprise instructions that cause the one or more processing devices to:

train a support vector machine by inputting documents labeled with the predetermined categories; and

wherein, to classify the documents selected by the user for viewing, the instructions include instructions that cause the one or more processing devices to use the trained support vector machine to label the documents with the predetermined categories.

21. The manufacture of claim 12, wherein the context is a time.

22. The manufacture of claim 12, wherein the context is a location.

23. A system for use in personalizing advertising for a user, the system comprising:

one or more processing devices; and

one or more storage devices storing instructions which, when executed, cause the one or more processing devices to implement;

a content server configured to present documents, from a set of documents, to a user such that the user can select documents for viewing;

at least one positive word vector formed using words contained in at least a segment of the documents in the set of documents that are selected by the user for viewing;

at least one negative word vector formed using words contained in at least a segment of the documents in the set of documents that are not selected by the user for viewing;

one or more document word vectors for at least some of the documents that were selected by the user for viewing;

a ranking engine configured to;

perform a vector space relationship analysis of the positive word vector, the negative word vector, and the document word vectors;

establish a document rank order of the documents selected by the user for viewing based on the performed vector space relationship analysis;

classify the documents selected by the user for viewing in predetermined categories; and

rank the predetermined categories based on the document rank order; and

an ad server, operatively coupled with the ranking engine, for presenting, to the user in a selected context, advertisements associated with the ranked categories.

24. The system of claim 23, wherein the documents in the set of documents are available at a website.

25. The system of claim 23, wherein the segment is one or more of a headline of the document, a lead of the document, or a summary of the document.

26. The system of claim 23, wherein the at least one negative word vector is generated by:

applying an expectation maximization algorithm to the documents in the document set that were not selected by the user for viewing to generate negative labels for certain ones of the non-selected documents, and positive labels for the rest of the non-selected documents; and

wherein generating at least one negative word vector comprises generating at least one negative word vector using words contained in at least a segment of the documents for which the expectation maximization algorithm generates negative labels.

27. The system of claim 23, wherein the instructions further cause the one or more processing devices to implement:

a document dictionary that contains dictionary words; and

wherein the document word vectors are generated by storing, for each document in the group of documents, a count of the number of occurrences of each dictionary word found in the document.

28. The system of claim 23, wherein the ranking engine is configured to perform the vector space relationship analysis by training a support vector machine with the positive word vector and the negative word vector and inputting the document word vectors into the trained support vector machine.

29. The system of claim 23, wherein the ranking engine is configured to perform the vector space relationship analysis by applying Rocchio'"'"'s method to the positive word vector, the negative word vector, and the document word vectors.

30. The system of claim 23, wherein the ranking engine is configured to perform the vector space relationship analysis by applying Naï

ve Bayes techniques to the positive word vector, the negative word vector, and the document word vectors.

31. The system of claim 23, wherein the instructions further cause the one or more processing devices to implement:

a support vector machine trained by receiving documents labeled with the predetermined categories; and

wherein the ranking engine is configured to classify the documents selected by the user for viewing by using the trained support vector machine to label the documents with the predetermined categories.

32. The system of claim 23, wherein the context is a time.

33. The system of claim 23, wherein the context is a location.