System and method for classification of microblog posts based on identification of topics

1. A method for assigning a topic to a collection of microblog posts, the method comprising:

receiving from at least one messaging service server, a plurality of posts, wherein:

each of the plurality of posts comprise post content, and

one or more of the posts comprise one or more links with an address to an external document;

causing a communications hardware component to initiate a communication via a communications network that accesses the external document at an external server that is associated with the address and fetches external content associated with the external document; and

by a generation module:

analyzing the post content included in each post to identify at least one label for that post,

for each post that includes a link, analyzing external content corresponding to the external document associated with that link to identify a topic, and

using a topic modeling technique to generate a trained topic model, wherein the trained topic model comprises, for each identified label, a plurality of topics and a plurality of words associated with each of the plurality of topics; and

saving the trained topic model to a computer-readable memory device,

wherein, the topic modeling technique comprises a User-Labeled Linked Hierarchical Dirichlet Process (L 2 -HDP) technique, that utilizes a user's interest variable vector to generate the trained topic model, wherein the user's interest variable vector is indicative of a topic composition of the set of documents associated with a user.

2. The method of claim 1 , further comprising:

receiving a new post associated with a user;

by a classifier module:

analyzing the new post by applying the trained topic model to a first plurality of words in the new post, and

generating a label for the new post; and

displaying the label, via a user interface.

3. The method of claim 1 , wherein:

each of the plurality of posts comprises a user identifier; and

wherein the method further comprises identifying and outputting one or more topics of interest for a single user by:

identifying a set of posts associated with the single user using an associated user identifier for the single user,

using the set of posts for generating a user-specific trained topic model such that the user-specific trained topic model is associated with the single user, and

using the user-specific trained topic model for identifying and outputting the one or more topics of interest for the single user.

4. The method of claim 1 , further comprising identifying the topic for each post by analyzing the content of that post to identify the topic.

5. The method of claim 1 , wherein the topic modeling technique comprises a User-Labeled Latent Dirichlet Allocation (uL-LDA) technique, that utilizes a user's interest variable vector to generate the trained topic model, wherein the user's interest variable vector is indicative of a topic composition of the set of documents associated with a user.

6. The method of claim 1 , wherein the external document comprises a plurality of labels, and the plurality of topics in the trained topic model comprise a set of topics shared amongst the plurality of labels.

7. The method of claim 1 , wherein the external document comprises a plurality of labels, and the L 2 -HDP analyzes the external content corresponding to the external document associated with that link if the plurality of labels comprises at least label similar to the identified at least one label.

8. The method of claim 1 , wherein the at least one label is selected from one of the following labels: a hashtag label, an emoticon-specific label, an @user label, a reply label, or a question label.

9. The method of claim 3 , further comprising updating the user-specific trained topic model for the single user after a pre-determined time.

10. A system for assigning a topic to a collection of microblog posts, the system comprising:

a processing device; and

a computer-readable memory containing programming instructions that, when executed by the processing device, cause the processing device to:

by an acquisition module, receive from at least one messaging service server, a plurality of posts, wherein:

each of the plurality of posts comprise post content, and

one or more of the posts comprise one or more links with an address to an external document;

by the acquisition module, cause a communications hardware component to initiate a communication via a communications network that accesses the external document at an external server that is associated with the address and fetches external content associated with the document;

by the generation module:

analyze the post content included in each post to identify at least one label for that post,

for each post that includes a link, analyze external content corresponding to the external document associated with that link to identify a topic, and

use a topic modeling technique to generate a trained topic model, wherein the trained topic model comprises, for each identified label, a plurality of topics and a plurality of words associated with each of the plurality of topics; and

by the generation module, save the trained topic model to a computer-readable memory device

wherein, the topic modeling technique comprises a User-Labeled Linked Hierarchical Dirichlet Process (L 2 -HDP) technique, that utilizes a user's interest variable vector to generate the trained topic model, wherein the user's interest variable vector is indicative of a topic composition of the set of documents associated with a user.

11. The system of claim 10 , further comprising programming instructions that, when executed by the processing device, cause the processing device to:

by the acquisition module, receive a new post associated with a user;

by a classifier module:

analyze the new post by applying the trained topic model to a first plurality of words in the new post, and

generate a label for the new post; and

display the label via a user interface.

12. The system of claim 10 , wherein each of the plurality of posts comprises a user identifier; and wherein the system further comprises programming instructions that, when executed by the processing device, cause the processing device to identify and output one or more topics of interest for a single user by:

identifying a set of posts associated with the single user using an associated user identifier for the single user,

using the set of posts for generating a user-specific trained topic model such that the user-specific trained topic model is associated with the single user, and

using the user-specific trained topic model for identifying and outputting the one or more topics of interest for the single user.

13. The system of claim 10 , further comprising programming instructions that, when executed by the processing device, cause the processing device to, by the generation module, identify the topic for each post by analyzing the content of that post to identify the topic.

14. The system of claim 10 , wherein the topic modeling technique comprises a User-Labeled Latent Dirichlet Allocation (uL-LDA) technique, that utilizes a user's interest variable vector to generate the trained topic model, wherein the user's interest variable vector is indicative of a topic composition of the set of documents associated with a user.

15. The system of claim 10 , wherein the external document comprises a plurality of labels, and the plurality of topics in the trained topic model comprise a set of topics shared amongst the plurality of labels.

16. The system of claim 10 , wherein the external document comprises a plurality of labels, and the L 2 -HDP analyzes the external content corresponding to the external document associated with that link if the plurality of labels comprises at least label similar to the identified at least one label.

17. The system of claim 10 , wherein the at least one label is selected from one of the following labels: a hashtag label, an emoticon-specific label, an @user label, a reply label, or a question label.

18. The system of claim 12 , further comprising programming instructions that, when executed by the processing device, cause the processing device to update the user-specific trained topic model for the single user after a pre-determined time.