Online thread retrieval using thread structure and query subjectivity

1. A method of handling queries for an online discussion forum, said method comprising:

receiving a query;

automatically classifying the query as subjective or objective;

thereupon calculating, for discussion threads of the query, at least one of:

a subjectivity score and an objectivity score;

said calculating comprising:

applying a maximum entropy model; and

incorporating, with respect to at least one discussion thread, at least one member taken from the group consisting of: a number of posts in a discussion thread; average number of words in posts, presence of a predetermined pattern among posts, a number of authors of posts within a discussion thread, average depth of each post, maximum depth of a post, and length of at least one reply; and

determining a degree of relevance to the query of: the discussion threads, and at least one post in the at least one discussion thread;

said determining of the degree of relevance of the query to the discussion threads comprising:

iteratively determining a relevance score with respect to each post in a discussion thread and then accepting a maximum relevance score with respect to a post in a discussion thread;

determining a penalty or reward regulizer with respect to choosing a predetermined number of posts for calculating a relevance score of a thread; and

including at least one of: a subjectivity score of the thread and an objectivity score of the thread; and

ranking the discussion threads based on said calculating and determining of a degree of relevance of the query to the discussion threads.

2. The method according to claim 1 , wherein said determining comprises determining a degree of relevance to the query of both of: the discussion threads, and at least one post in the discussion threads.

3. The method according to claim 2 , wherein said determining of a degree of relevance of the query to the discussion threads comprises collecting top relevant threads and determining a relevance to the query of the top relevant threads.

4. The method according to claim 2 , wherein said determining of a degree of relevance of the query to at least one post comprises collecting top relevant threads and determining a relevance to the query of all posts in the top relevant threads.

5. The method according to claim 1 , comprising applying learned balancing factors to each of: the relevance score with respect to a post in a discussion thread, the penalty or reward regulizer, and the at least one of: a subjectivity score of the thread and an objectivity score of the thread.

6. The method according to claim 5 , wherein the

penalty or reward regulizer comprises a calculation of graph cover with respect to the chosen number of posts for calculating a relevance score of a thread.

7. The method according to claim 1 , wherein said determining of a relevance score with respect to each post in a discussion thread comprises employing a query likelihood model with smoothing.

8. The method according to claim 7 , wherein said employing of a query likelihood model with smoothing comprises determining an average relevance of scorewise top-k posts in a thread.

9. The method according to claim 1 wherein said calculating comprises calculating a subjectivity score of discussion threads of the query upon classifying the query as subjective.

10. The method according to claim 1 wherein said calculating comprises calculating an objectivity score of discussion threads of the query upon classifying the query as objective.

11. The method according to claim 1 , wherein said calculating comprises incorporating at least one factor comprising a number of posts in a discussion thread.

12. The method according to claim 1 , wherein said calculating comprises incorporating, with respect to at least one discussion thread, at least one member taken from the group consisting of: average number of words in posts, presence of a predetermined pattern among posts, a number of authors of posts within a discussion thread, average depth of each post, maximum depth of a post, and length of at least one reply.

13. The method according to claim 11 , wherein said applying of a maximum entropy model comprises employing kernel density estimation to convert non-monotonic features to monotonic features in the maximum entropy model.

14. An apparatus comprising:

at least one processor; and

a computer readable storage medium having computer readable program code embodied therewith and executable by the at least one processor, the computer readable program code comprising:

computer readable program code configured to receive a query;

computer readable program code configured to automatically classify the query as subjective or objective;

computer readable program code configured, to

thereupon calculate, for discussion threads of the query, at least one of: a subjectivity score and an objectivity score;

wherein to calculate comprises:

applying a maximum entropy model; and

incorporating, with respect to at least one discussion thread, at least one member taken from the group consisting of: a number of posts in a discussion thread; average number of words in posts, presence of a predetermined pattern among posts, a number of authors of posts within a discussion thread, average depth of each post, maximum depth of a post, and length of at least one reply; and

computer readable program code configured to determine a degree of relevance to the query of at least one of: the discussion threads, and at least one post in the at least one discussion thread;

wherein to determine the degree of relevance of the query to the discussion threads comprises:

iteratively determining a relevance score with respect to each post in a discussion thread and then accenting a maximum relevance score with respect to a post in a discussion thread;

determining a penalty or reward regulizer with respect to choosing a predetermined number of posts for calculating a relevance score of a thread; and

including at least one of: a subjectivity score of the thread and an objectivity score of the thread; and

computer readable program code configured to rank the discussion threads based on the calculating and determining of a degree of relevance of the query to the discussion threads.

15. A computer program product comprising:

at least one processor; and

a computer readable storage medium having computer readable program code embodied therewith and executable by the at least one processor, the computer readable program code comprising:

computer readable program code configured to receive a query;

computer readable program code configured to automatically classify the query as subjective or objective;

computer readable program code configured, to

thereupon calculate, for discussion threads of the query, at least one of: a subjectivity score and an objectivity score;

wherein to calculate comprises:

applying a maximum entropy model; and

incorporating, with respect to at least one discussion thread, at least one member taken from the group consisting of: a number of posts in a discussion thread; average number of words in posts, presence of a predetermined pattern among posts, a number of authors of posts within a discussion thread, average depth of each post, maximum depth of a post, and length of at least one reply; and

computer readable program code configured to determine a degree of relevance to the query of at least one of: the discussion threads, and at least one post in the at least one discussion thread;

wherein to determine the degree of relevance of the query to the discussion threads comprises:

iteratively determining a relevance score with respect to each post in a discussion thread and then accepting a maximum relevance score with respect to a post in a discussion thread;

determining a penalty or reward regulizer with respect to choosing a predetermined number of posts for calculating a relevance score of a thread; and

including at least one of: a subjectivity score of the thread and an objectivity score of the thread; and

computer readable program code configured to rank the discussion threads based on the calculating and determining of a degree of relevance of the query to the discussion threads.

16. The computer program product according to claim 15 , wherein said computer readable program code is configured to determine a degree of relevance to the query of both of: the discussion threads, and at least one post in the discussion threads.