Ranking relevant discussion groups
Messages are collected and processed to determine topic identifiers that correspond to discussion groups. Queries are received and multiple discussion groups that are relevant to the query are determined based on the messages that are associated with the discussion groups and the topic identifiers associated with the discussion groups. The relevant discussion groups are ranked using a group preference model that simulates the behavior of a hypothetical seeker that considers discussion groups by selecting a message author who is an authority in a particular group, and exploring the discussion groups that are preferred by the selected author. The behavior of the seeker is simulated using a stationary Markov process and is used to generate a probability distribution that is used to rank the relevant discussion groups. The ranked relevant discussion groups are provided in response to the query.
1. A method comprising:
receiving a query by a computing device;
determining a plurality of discussion groups that are relevant to the query by the computing device, wherein each discussion group is associated with a plurality of messages and each message is associated with an author;
for each discussion group of the plurality of discussion groups, determining an authority score for each author associated with a message of the discussion group by the computing device;
for each discussion group of the plurality of discussion groups, determining a teleport score for the discussion group, including:
determining a total number of messages associated with the discussion group;
determining a number of messages associated with the discussion group that are relevant to the query; and
determining the teleport score based on a ratio of messages that are relevant to the query to the total number of messages associated with the discussion group;
for each author associated with a message, determining a preference score for the author for each discussion group of the plurality of discussion groups by the computing device, wherein determining the preference score comprises determining the preference score based on a proportion of a number of occurrences of the discussion group that the author participated in to a total number of occurrences of the plurality of discussion groups that the author participated in; and
ranking the discussion groups of the plurality of discussion groups using the preference scores, the teleport scores, and the authority scores by the computing device, wherein using the preference scores, the teleport scores, and the authority scores comprises generating a distribution for the query and the discussion groups using the preference scores, the teleport scores, and the authority scores, and wherein ranking the discussion groups comprises ranking the discussion groups using the generated distribution.
2. The method of claim 1 , further comprising providing the ranked discussion groups in response to the query.
3. The method of claim 1 , wherein determining an authority score for an author comprises determining a number of followers associated with the author and determining the authority score based on the determined number of followers.
4. The method of claim 1 , wherein determining an authority score for an author comprises determining a number of times that the author is referenced in a message of the plurality of messages and determining the authority score based on the determined number of times.
5. The method of claim 1 , wherein determining an authority score for an author comprises determining a number of direct replies associated with the author and determining the authority score based on the determined number of direct replies.
6. The method of claim 1 , wherein the authority score for an author may be determined based on one or more of noun frequency or mention weights.
7. The method of claim 1 , wherein determining a preference score for an author for a discussion group comprises:
determining a number of occurrences of the discussion group that the author participated in; and
determining the preference score based on the determined number.
8. The method of claim 1 , wherein determining a preference score for an author comprises determined a transition probability matrix and determining the preference score based on the transition probability matrix.
9. The method of claim 1 , wherein each discussion group is associated with a topic identifier and each message associated with a discussion group includes the topic identifier associated with the discussion group.
10. The method of claim 1 , further comprising:
determining the number of occurrences of the discussion group that the author participated in; and
determining the total number of occurrences of the plurality of discussion groups that the author participated in.
11. A method comprising:
receiving a plurality of messages at a computing device, wherein each message includes a topic identifier and is associated with an author;
determining topic identifiers of the plurality of messages that represent discussion groups by the computing device;
receiving a query at the computing device;
determining topic identifiers of the determined topic identifiers that are relevant to the received query by the computing device;
ranking the determined topic identifiers based on the messages that include the determined topic identifiers by the computing device, wherein ranking the determined topic identifiers comprises:
for each determined topic identifier, determining an authority score for each author associated with a message that includes the topic identifier;
for each determined topic identifier, determining a teleport score, including:
determining a total number of messages associated with the topic identifier;
determining a number of messages associated with the topic identifier that are relevant to the query; and
determining the teleport score based on a ratio of messages that are relevant to the query to the total number of messages associated with the topic identifier;
for each author associated with a message, determining a preference score for the author for each determined topic identifier using the messages associated with the author, wherein determining the preference score comprises determining the preference score based on a proportion of a number of occurrences of the discussion group that the author participated in to a total number of occurrences of the plurality of discussion groups that the author participated in; and
ranking the determined topic identifiers using the preference scores, the teleport scores, and the authority scores, wherein using the preference scores, the teleport scores, and the authority scores comprises generating a distribution for the query and the discussion groups using the preference scores and the authority scores, and wherein ranking the discussion groups comprises ranking the discussion groups using the generated distribution; and
providing the ranked determined topic identifiers in response to the query by the computing device.
12. The method of claim 11 , wherein a topic identifier represents a discussion group if the messages that include the topic identifier are one or more of periodic, synchronous, or cohesive.
13. The method of claim 11 , wherein determining an authority score for an author comprises determining a number of followers associated with the author and determining the authority score based on the determined number of followers.
14. The method of claim 11 , wherein determining a preference score for an author for a topic identifier comprises:
determining a number of occurrences of the discussion group represented by the topic identifier that the author participated in; and
determining the preference score based on the determined number.
15. A system comprising:
at least one computing device; and
a discussion group engine adapted to:
receive a query and a plurality of discussion groups that are relevant to the query, wherein each discussion group is associated with a plurality of messages and each message is associated with an author;
for each discussion group of the plurality of discussion groups, determine an authority score for each author associated with a message associated with the discussion group;
for each author associated with a message, determine a preference score for the author for each discussion group of the plurality of discussion groups, wherein determining the preference score comprises determining the preference score based on a proportion of a number of occurrences of the discussion group that the author participated in to a total number of occurrences of the plurality of discussion groups that the author participated in;
for each group of the plurality of discussion groups, determine a teleport score, including:
determining a total number of messages associated with the discussion group;
determining a number of messages associated with the discussion group that are relevant to the query; and
determining the teleport score based on a ratio of messages that are relevant to the query to the total number of messages associated with the discussion group; and
rank the discussion groups of the plurality of discussion groups using the preference scores, the authority scores, and the teleport scores, wherein using the preference scores, the authority scores, and the teleport scores comprises generating a distribution for the query and the discussion groups using the preference scores, the authority scores, and the teleport scores, and wherein ranking the discussion groups comprises ranking the discussion groups using the generated distribution.
16. The system of claim 15 , wherein the discussion group engine is further adapted to provide the ranked discussion groups in response to the received query.