A Semi - supervised Text Clustering Algorithm Based on Pairwise Constraints ★

In this paper, an active learning method which can effectively select pairwise constraints during clustering procedure was presented. A novel semi-supervised text clustering algorithm was proposed, which employed an effective pairwise constraints selection method. As the samples on the fuzzy boundary are far away from the cluster center in the clustering procedure, they can be easily divided into the wrong clusters. Therefore, we choose the pairwise constraint points from the fuzzy boundary to guide the clustering process towards appropriate partition. The experimental results show that the proposed algorithm can effectively improve the text clustering results by using the same amount of pairwise constraints.