Bridging Social Silos for Knowledge Discovery and Sharing

• US 20120296967A1
• Filed: 05/20/2011
• Published: 11/22/2012
• Est. Priority Date: 05/20/2011
• Status: Active Grant

• Current Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Original Assignee: Cisco Technology, Inc. (Cisco Systems, Inc.)
• Inventors: Tao, Yicheng, Bochinski, Stephen Daniel, Sambhus, Sonali Mihir
• Granted Patent: US 8,793,312 B2
• Time in Patent Office: Days
• US Class Current: 709/204
• CPC Class Codes: G06Q 30/0251 Targeted advertisements

Abstract

Techniques are provided herein for obtaining network usage information about a plurality of users, analyzing the network usage information to detect a plurality of subjects and determine a set of relationship scores describing the strength of the relationship between users and subjects, and analyzing the sets of relationship scores to detect a plurality of virtual communities formed among the users. The virtual communities are used to detect subjects to recommend to members of the virtual communities, such as topics or content items that other users of the community have found of interest.

Claims

1. A method comprising:

obtaining, by a computer, network usage information about a user;

detecting, by the computer, a plurality of subjects by analyzing the network usage information;

analyzing, by the computer, the network usage information of the user to determine a set of relationship scores, wherein each relationship score describes the strength of a relationship between the user and an individual subject in the plurality of subjects;

analyzing, by the computer, the set of relationship scores and relationship scores of other users to detect a plurality of virtual communities, wherein the user may be a member of more than one virtual community; and

recommending, by the computer, one or more content items stored in a storage area to the user based on the virtual communities of which the user is a member.

2. The method of claim 1, wherein the subjects are content items.

3. The method of claim 1, further comprising:

updating virtual community membership by the computer repeating said obtaining, detecting, analyzing the network usage information and analyzing the set of relationship scores steps, wherein said updating may result in changes in one or more relationship scores for the user, and the user being added or deleted as a member of one or more virtual communities.

4. The method of claim 3, wherein said updating occurs at predetermined time intervals.

5. The method of claim 1, wherein said analyzing the set of relationship scores further comprises:

applying, by the computer, a statistical correlation model to the set of relationship scores and the relationship scores of the other users in order to determine which of the other users are statistically correlated with the user; and

assigning, by the computer, the user to the same virtual community as the other users who are statistically correlated with the user.

6. The method of claim 5, wherein the statistical correlation model is a similarity model or a k-nearest neighbors model.

7. The method of claim 5, wherein said recommendation further comprises, for each virtual community of which a user is a member:

determining, by the computer, one or more subjects that have the highest relationship scores to other members of the same virtual community; and

using the determined subjects to identify one or more content items stored in the storage area and associated with the determined subjects to recommend to the user.

8. The method of claim 1, wherein the subjects are topics.

9. The method of claim 8, wherein said analyzing the set of relationship scores further comprises the computer:

generating a user-topic matrix comprising the set of relationship scores for the user and the relationship scores of the other users, wherein the user-topic matrix identifies one or more topics which are most strongly related with the user and each of the other users,

applying a topic model to the user-topic matrix in order to generate a topic-community matrix and a user-community matrix,

wherein the topic-community matrix identifies for each topic one or more virtual communities which are most strongly associated with the topic, and

wherein the user-community matrix identifies one or more virtual communities which are most strongly associated with the user, and in which the user is considered a member.

10. The method of claim 9, wherein the topic model is a latent Direchlet allocation model.

11. The method of claim 9, wherein said recommendation further comprises:

determining, by the computer, one or more topics that have the highest relationship scores for the user;

for each determined topic, the computer;

using the topic-community matrix to identify one or more virtual communities which are most strongly associated with the topic;

for each identified virtual community, using the user-community matrix to identify one or more other users who are most strongly associated with that virtual community;

for each identified other user, using the determined topic to search the network usage information about the identified other user in order to find content items to recommend to the user receiving a recommendation.

12. An apparatus comprising:

a memory having a plurality of content items stored therein; and

a processor configured to;

obtain network usage information about a user;

detect a plurality of subjects by analyzing the network usage information;

analyze the network usage information of the user to determine a set of relationship scores, wherein each relationship score describes the strength of a relationship between the user and an individual subject in the plurality of subjects;

analyze the set of relationship scores and the relationship scores of other users to detect a plurality of virtual communities, wherein the user may be a member of more than one virtual community; and

recommend one or more content items stored in said memory to the user based on the virtual communities of which the user is a member.

13. The apparatus of claim 12, wherein said analyzing the set of relationship scores comprises the processor being further configured to:

apply a statistical correlation model to the set of relationship scores and the relationship scores of the other users in order to determine which of the other users are statistically correlated with the user, wherein the statistical correlation model is a similarity model or a k-nearest neighbors model; and

assign the user to the same virtual community as the other users who are statistically correlated with the user; and

wherein said recommendation comprises, for each virtual community of which the user is a member, the processor being further configured to;

determine one or more subjects that have the highest relationship scores to other members of the same virtual community; and

use the determined subjects to identify one or more content items associated with the determined subjects to recommend to the user.

14. The apparatus of claim 12, wherein the subjects are topics, and wherein said analyzing the set of relationship scores comprises the processor being further configured to:

generate a user-topic matrix comprising the set of relationship scores for the user and the relationship scores of the other users, wherein the user-topic matrix identifies one or more topics which are most strongly related with the user and each of the other users,

apply a latent Direchlet allocation model to the user-topic matrix in order to generate a topic-community matrix and a user-community matrix,

wherein the topic-community matrix identifies for each topic one or more virtual communities which are most strongly associated with the topic, and

wherein the user-community matrix identifies one or more virtual communities which are most strongly associated with the user, and in which the user is considered a member.

15. The apparatus of claim 14, wherein said recommendation further comprises the processor being further configured to:

determine one or more topics that have the highest relationship scores for the user;

for each determined topic;

use the topic-community matrix to identify one or more virtual communities which are most strongly associated with the topic;

for each identified virtual community, use the user-community matrix to identify one or more other users who are most strongly associated with that virtual community;

for each identified other user, use the determined topic to search the network usage information about the identified other user in order to find content items to recommend to the user receiving a recommendation.

16. One or more computer readable media encoded with instructions that, when executed by a processor, cause the processor to:

obtain network usage information about a user;

detect a plurality of subjects by analyzing the network usage information;

analyze the network usage information of the user to determine a set of relationship scores, wherein each relationship score describes the strength of a relationship between the user and an individual subject in the plurality of subjects;

analyze the set of relationship scores and the relationship scores of other users to detect a plurality of virtual communities, wherein the user may be a member of more than one virtual community; and

recommend one or more content items stored in said memory to the user based on the virtual communities of which the user is a member.

17. The computer readable media of claim 16, further comprising instructions that when executed cause the processor to:

update virtual community membership by the computer repeating said obtaining, detecting, analyzing the network usage information and analyzing the set of relationship scores steps, wherein said updating may result in changes in one or more relationship scores for the user, and the user being added or deleted as a member of one or more virtual communities.

18. The computer readable media of claim 16, wherein said analyzing the set of relationship scores comprises further instructions that when executed cause the processor to:

apply a statistical correlation model to the set of relationship scores and the relationship scores of the other users in order to determine which of the other users are statistically correlated with the user, wherein the statistical correlation model is a similarity model or a k-nearest neighbors model; and

assign the user to the same virtual community as the other users who are statistically correlated with the user.

19. The computer readable media of claim 16, wherein said recommendation further comprises, for each virtual community of which a user is a member, instructions that when executed cause the processor to:

determine one or more subjects that have the highest relationship scores to other members of the same virtual community; and

use the determined subjects to identify one or more content items associated with the determined subjects to recommend to the user.

20. The computer readable media of claim 16, wherein the subjects are topics, and wherein said analyzing the set of relationship scores comprises further instructions that when executed cause the processor to:

generate a user-topic matrix comprising the set of relationship scores for the user and the relationship scores of the other users, wherein the user-topic matrix identifies one or more topics which are most strongly related with the user and each of the other users,

apply a latent Direchlet allocation model to the user-topic matrix in order to generate a topic-community matrix and a user-community matrix,

wherein the topic-community matrix identifies for each topic one or more virtual communities which are most strongly associated with the topic, and

wherein the user-community matrix identifies one or more virtual communities which are most strongly associated with the user, and in which the user is considered a member.