Affinity propagation clustering on oral conversation texts

Abstract

This article describes a method that applied the new clustering algorithm Affinity Propagation (AP) on oral conversation texts. And we used various measures of similarity to test the performance of this new algorithm. In our experiment, we compared the AP with the Self-Organizing Map (SOM) which is a kind of classical clustering algorithm. The experimental results showed us the Kullback-Leibler Divergence (Relative Entropy) is the best choice in affinity propagation algorithm, and it produced a better result than SOM