Clustering with similarity-adjusted entropy

A method for clustering the authors of messages from a website for online discussion, comprising the operations of:

• initializing the method by assigning the messages to a first logical cluster that is one of a plurality of logical clusters, wherein each message is associated with an author, a category in a domain, and a rating based on feedback from readers of the message;
• determining a first measure of similarity-adjusted entropy for the plurality of logical clusters, wherein similarity-adjusted entropy is based at least in part on the category and the rating associated with each of the messages;
• randomly reassigning one of the messages to a second logical cluster, wherein the second logical cluster is randomly selected from the plurality of logical clusters;
• determining a second measure of similarity-adjusted entropy for the plurality of logical clusters;
• undoing the reassignment, if the second measure is greater than or equal to the first measure;
• replacing the first measure with the second measure, if the second measure is less than the first measure;
• and
• repeating all of the operations of the method following initialization until they converge, wherein each operation of the method is executed by a processor.