Automatic profiling of social media users

1. A method, comprising:

obtaining a plurality of messages posted by a user via a microblogging service;

determining a set of prototypical replied accounts associated with a particular class or prototypical retweeted accounts associated with the particular class, wherein determining includes identifying accounts cited in messages of users of the particular class, the messages of the users of the particular class being replies or retweets;

obtaining a set of feature values associated with the user, each of the set of feature values corresponding to a different one of a set of one or more features, wherein at least a portion of the set of feature values is obtained based, at least in part, on content of the plurality of messages posted by the user via the microblogging service, wherein at least one of the portion of the set of feature values is determined based, at least in part, upon activity of the user with respect to the set of prototypical replied accounts associated with the particular class or prototypical retweeted accounts associated with the particular class; and

classifying by a processor the user based upon the set of feature values associated with the user such that the user is classified based, at least in part, on the content of the plurality of messages, wherein classifying the user includes labeling the user to indicate whether the user is a member of the particular class;

wherein one or more of the portion of the set of feature values indicates a sentiment of the user with respect to one or more prototypical words of a set of prototypical words representative of the particular class, the set of prototypical words representing the particular class including a plurality of prototypical words, the sentiment being positive, negative, or neutral.

2. The method as recited in claim 1 , wherein one or more of the portion of the set of feature values is based, at least in part, upon social connections established by the user via the microblogging service with respect to a set of prototypical friend accounts associated with the particular class.

3. The method as recited in claim 2 , wherein the one or more of the portion of the set of feature values include a feature value that indicates whether the user is a follower of the set of prototypical friend accounts associated with the particular class.

4. The method as recited in claim 1 , wherein the set of features comprises a plurality of features, and wherein classifying by a processor the user based upon the set of feature values associated with the user such that the user is classified based, at least in part, on the content of the plurality of messages comprises:

applying a machine learned mathematical model to the set of feature values associated with the user, wherein the machine learned model includes a weight associated with each of the plurality of features.

5. The method as recited in claim 4 , further comprising:

training the machine learned mathematical model using a set of seed users for which class membership is known.

6. The method as recited in claim 1 , wherein classifying the user further comprises:

ascertaining whether the user is a member of the particular class.

7. The method as recited in claim 1 , wherein one or more of the set of feature values is based, at least in part, upon social connections established by the user via the microblogging service.

8. The method as recited in claim 1 , wherein one or more of the set of feature values is derived based upon activity of the user with respect to one or more social connections of the user that have been established via the microblogging service, wherein the activity of the user includes at least one of: following the one or more social connections of the user, replying to messages of the one or more social connections of the user, or reposting messages of the one or more social connections of the user.

9. The method as recited in claim 1 , wherein one or more of the portion of the set of feature values is based, at least in part, upon a subset of the set of prototypical words that are identified within the plurality of messages, wherein classifying the user comprises:

ascertaining whether the user is a member of the particular class.

10. The method as recited in claim 1 , wherein one or more of the portion of the set of feature values is based, at least in part, upon a subset of a set of prototypical hashtags that are identified in the plurality of messages, the set of prototypical hashtags representing the particular class, wherein classifying the user comprises:

ascertaining whether the user is a member of the particular class.

11. The method as recited in claim 1 , wherein classifying the user comprises detecting a political affiliation of the user, performing ethnicity identification to identify an ethnicity of the user, detecting a gender of the user, or detecting affinity of the user for a particular business.

12. The method as recited in claim 1 , wherein classifying the user comprises:

representing the user by a multinomial distribution over a plurality of topics.

13. The method as recited in claim 1 , wherein the particular class is a specific class of users.

14. The method as recited in claim 1 , wherein one or more of the portion of the set of feature values is based, at least in part, upon a subset of a set of prototypical topics that are identified in the plurality of messages, the set of prototypical topics representing the particular class, wherein classifying the user comprises:

ascertaining whether the user is a member of the particular class.

15. A non-transitory computer-readable storage medium storing thereon computer-readable instructions, comprising:

instructions for obtaining a plurality of messages posted by a user via a microblogging service;

instructions for determining a set of prototypical replied accounts associated with a particular class or prototypical retweeted accounts associated with the particular class, wherein determining includes identifying accounts cited in messages of users of the particular class, the messages of the users of the particular class being replies or retweets;

instructions for obtaining a set of feature values associated with the user, each of the set of feature values corresponding to a different one of a set of one or more features, wherein at least a portion of the set of feature values is obtained based, at least in part, on content of the plurality of messages posted by the user via the microblogging service, wherein at least one of the portion of the set of feature values is determined based, at least in part, upon activity of the user with respect to the set of prototypical replied accounts associated with the particular class or prototypical retweeted accounts associated with the particular class; and

instructions for classifying the user based upon the set of feature values associated with the user such that the user is classified based, at least in part, on the content of the plurality of messages, wherein classifying the user includes labeling the user to indicate whether the user is a member of the particular class;

wherein one or more of the portion of the set of feature values indicates a sentiment of the user with respect to one or more prototypical words of a set of prototypical words representative of the particular class, the set of prototypical words representing the particular class including a plurality of prototypical words, the sentiment being positive, negative, or neutral.

16. The non-transitory computer-readable storage medium as recited in claim 15 , further comprising:

instructions for aggregating the plurality of messages transmitted or posted by the user via the microblogging service into a single document;

wherein obtaining the set of feature values associated with the user comprises deriving the at least a portion of the set of feature values, at least in part, from at least a portion of the content of the plurality of messages in the single document.

17. The non-transitory computer-readable storage medium as recited in claim 16 , wherein the instructions for obtaining a set of feature values associated with the user comprises:

instructions for generating a set of one or more numerical values characterizing linguistic content of the plurality of messages of the user.

18. The non-transitory computer-readable storage medium as recited in claim 16 , wherein the set of feature values pertains to at least one of the set of prototypical words representing the particular class, a set of prototypical topics representing the particular class, or a set of prototypical hashtags representing the particular class, wherein classifying the user comprises ascertaining whether the user is a member of the particular class.

19. The non-transitory computer-readable storage medium as recited in claim 16 , wherein one of the plurality of messages includes an audio message, wherein aggregating the plurality of messages transmitted or posted by the user via the microblogging service into a single document comprises:

converting the audio message into text.

20. The non-transitory computer-readable storage medium as recited in claim 15 , wherein one or more of the set of feature values pertains to one or more topics of interest to the user or lexical usage of the user within the plurality of messages.

21. An apparatus, comprising:

a processor; and

a memory, at least one of the processor or the memory being adapted for:

obtaining a plurality of messages posted by a user via a microblogging service;

determining a set of prototypical replied accounts associated with a particular class or prototypical retweeted accounts associated with the particular class, wherein determining includes identifying accounts cited in messages of users of the particular class, the messages of the users of the particular class being replies or retweets;

obtaining a set of feature values associated with the user, each of the set of feature values corresponding to a different one of a set of one or more features, wherein at least a portion of the set of feature values is obtained based, at least in part, on content of the plurality of messages posted by the user, wherein at least one of the portion of the set of feature values is determined based, at least in part, upon activity of the user with respect to the set of prototypical replied accounts associated with the particular class or prototypical retweeted accounts associated with the particular class; and

classifying the user based upon the set of feature values associated with the user such that the user is classified based, at least in part, on the content of the plurality of messages, wherein classifying the user includes labeling the user to indicate whether the user is a member of the particular class;

wherein one or more of the portion of the set of feature values indicates a sentiment of the user with respect to one or more prototypical words of a set of prototypical words that is representative of the particular class, the set of prototypical words representative of the particular class including a plurality of prototypical words, the sentiment being positive, negative, or neutral.

22. The apparatus as recited in claim 21 , wherein classifying the user further comprises:

ascertaining whether the user is a member of the particular class.

23. The apparatus as recited in claim 21 , wherein one or more of the set of feature values is based, at least in part, upon a subset of the set of prototypical words identified within the plurality of messages, wherein classifying the user comprises:

ascertaining whether the user is a member of the particular class.

24. The apparatus as recited in claim 21 , further comprising:

ascertaining whether the user is a member of the particular class.

25. The apparatus as recited in claim 21 , wherein one or more of the set of feature values is based, at least in part, upon a subset of a set of prototypical hashtags that are identified in the plurality of messages, the set of prototypical hashtags representing the particular class, wherein classifying the user comprises:

ascertaining whether the user is a member of the particular class.

26. The apparatus as recited in claim 21 , wherein each one of a plurality of classes is associated with a corresponding set of topics about which members of the one of the plurality of classes are most likely to communicate, wherein each set of topics is a subset of a plurality of topics, wherein obtaining a set of feature values associated with the user comprises:

identifying one or more of the plurality of topics from linguistic content of the plurality of messages, thereby enabling the user to be classified in one or more of the plurality of classes.

27. The apparatus as recited in claim 21 , wherein one or more of the set of feature values is based, at least in part, upon a subset of a subset of a set of prototypical topics that are identified in the plurality of messages, the set of prototypical topics representing the particular class, wherein classifying the user comprises:

ascertaining whether the user is a member of the particular class.