A Model Framework for Enhancing Document Clustering Through Side Information

Abstract

Mining textual data is the need of this era. In most text mining applications, side information is associated with the text documents. This side information consists of document origin information, links present in document, user-access behavior which can be retrieved from web logs, and different kinds of non-textual attributes. Such side information may include explanatory data which can be beneficial for text clustering. Conventional text clustering methods are available that perform quite well but do not consider such attributes. As side information consists of meaningful data, it can assist in enriching the quality of clusters by incorporating such side information into clustering process. Nevertheless, not all side information is important. Hence it should be incorporated carefully. Herein, an effective clustering technique is proposed which adds only important side information in the clustering process and determines if incorporating side information improves the cluster quality. The clustering technique is extended to classification problem.